Type-specific detection and quantification of circulating tumor HPV DNA by digital PCR in surgically resected cervical cancer

Abstract

Objective

Circulating tumor human papillomavirus (HPV) DNA (ctHPV DNA) has shown clinical utility in advanced cervical cancer; however, its role in surgically resected disease remains unclear. We evaluated the detectability and clinicopathological associations of ctHPV DNA in preoperative plasma from patients with surgically resected cervical cancer using digital polymerase chain reaction (dPCR).

Methods

We analyzed 49 surgically resected cervical cancer patients. HPV genotyping of formalin-fixed paraffin-embedded tumor tissues was performed using both PGMY09/11 consensus PCR and novel genotype-specific dPCR assays targeting 6 high-risk HPV types (16, 18, 31, 39, 45, and 52). Plasma cell-free DNA was analyzed using dPCR for HPV genotypes. Associations between ctHPV DNA copy number and clinicopathological features were evaluated.

Results

HPV genotypes were identified in 42 of 49 tumor tissues (85.7%), with dPCR showing superior sensitivity (95.8%) compared to the PGMY method (77.6%). ctHPV DNA was detected in 12 of 42 evaluable plasma samples (28.6%), and plasma genotypes were concordant with those of the corresponding tumor tissues. ctHPV DNA copy numbers were significantly higher in cases with lymphovascular invasion (p=0.034) and lymph node metastasis (p=0.010), but were not correlated with tumor size, stage, or histologic type. ctHPV DNA was more consistently detectable than PIK3CA hotspot mutations in plasma, supporting its robustness as a biomarker in liquid biopsy settings.

Conclusion

This study highlights ctHPV DNA as a promising noninvasive biomarker in surgically resected cervical cancer. Its association with lymphovascular invasion and lymph node metastasis supports its potential role in preoperative risk stratification and personalized surgical planning.

Synopsis

Circulating tumor human papillomavirus (ctHPV) DNA is detectable in certain surgically resected cervical cancer using digital PCR (dPCR). dPCR shows higher sensitivity than conventional genotyping. Detection correlates with lymphovascular invasion and lymph node metastasis. ctHPV DNA may guide preoperative risk stratification and surgical planning.

INTRODUCTION

Cervical cancer remains a major global health burden as the fourth most common cancer in women worldwide [1]. Persistent infection with high-risk human papillomavirus (HPV), particularly HPV16 and HPV18, comprises over 70% of cases [2]. Although most HPV infections are transient and spontaneously cleared, persistent infection may result in viral genome integration into the host DNA [3], leading to sustained expression of the viral oncogenes E6 and E7 [4,5]. These proteins inactivate tumor suppressors p53 and pRb, promoting genomic instability and progression from high-grade intraepithelial neoplasia to invasive cancer, often accompanied by lymphovascular invasion and lymph node metastasis [3,4,5,6]. The introduction of prophylactic HPV vaccines has markedly reduced cervical cancer incidence in many high-income countries. However, vaccination coverage in Japan has remained low for the past decade owing to the suspension of proactive governmental recommendations, and the incidence and mortality of cervical cancer remain higher than expected for a high-income nation [2,7]. Thus, more effective diagnostic tools are warranted to refine risk stratification and expand individualized treatment strategies.

HPV genotype plays a key role in carcinogenesis, clinical behavior, and prognosis of cervical cancer. HPV16 is most associated with squamous cell carcinoma (SCC), whereas HPV18 is more strongly linked to adenocarcinoma, although both types occur across histological subtypes [2,8]. In Japan, 8 high-risk HPV types (HPV16, 18, 31, 33, 35, 45, 52, and 58) account for approximately 90% of cervical intraepithelial lesion grade 2 or higher lesions and over 94% of invasive cervical cancer [2,9]. These findings emphasize the clinical and epidemiological relevance of genotype-specific detection methods.

Although cervical cancer diagnosis is confirmed histologically, preoperative risk stratification largely relies on imaging and serum tumor markers, such as SCC antigen, carcinoembryonic antigen, and cancer antigen 125. However, these tools have limited sensitivity for detecting microscopic disease and predicting pathological features, such as lymph node metastasis, particularly in early-stage cases [10]. Fertility-sparing surgery, such as radical trachelectomy, is generally limited to lymph node–negative patients; if nodal involvement is identified intraoperatively, the disease is upstaged to International Federation of Gynecology and Obstetrics (FIGO) IIIC1, and conversion to hysterectomy is usually required. When nodal metastasis is detected postoperatively, adjuvant chemoradiotherapy is usually indicated. Thus, reliable preoperative markers that can indicate the risk of lymph node metastasis would be of substantial clinical value.

Liquid biopsy, particularly the detection of circulating tumor DNA (ctDNA), has emerged as a minimally invasive approach for molecular profiling of solid tumors. In several malignancies, such as breast, lung, and colorectal cancers, ctDNA has demonstrated clinical utility for diagnosis, disease monitoring, and prognostication [11,12]. However, the application of mutation-based ctDNA assays in cervical cancer has been limited by the relatively low prevalence of recurrent driver mutations. Large-scale genomic studies, including The Cancer Genome Atlas and Japanese C-CAT database, have shown that PIK3CA alterations are the most frequent events in SCC, whereas adenocarcinomas may also harbor KRAS mutations. However, no single hotspot mutation is consistently shared across cases [8,13]. This genetic heterogeneity restricts the sensitivity of conventional mutation-based ctDNA detection methods.

An alternative and potentially more effective approach is the detection of circulating tumor HPV DNA (ctHPV DNA), which directly reflects the viral etiology of HPV-driven tumors. ctHPV DNA comprises viral fragments released from tumor cells harboring integrated or episomal HPV genomes. In oropharyngeal SCC, ctHPV DNA correlates with disease burden and recurrence risk [14]. In cervical cancer, studies of advanced disease have demonstrated that ctHPV DNA detection is associated with tumor size, FIGO stage, nodal status, and recurrence [15,16,17]. These results suggest that ctHPV DNA may represent a tumor-specific, noninvasive biomarker of clinical relevance. Despite these promising findings, the clinical significance of ctHPV DNA in surgically resected cervical cancer remains unelucidated. Key unanswered questions include whether ctHPV DNA can be reliably detected in localized tumors and whether its quantitative level is associated with high-risk pathological features, such as lymph node metastasis.

In this study, we evaluated the detectability and quantitative levels of ctHPV DNA in preoperative plasma from patients with surgically resected cervical cancer using digital polymerase chain reaction (dPCR), and assessed its association with key clinicopathological parameters to determine its potential as a noninvasive biomarker.

MATERIALS AND METHODS

1. Study population

We included 49 patients diagnosed with cervical cancer who underwent primary surgical treatment (radical hysterectomy, radical trachelectomy, or modified radical hysterectomy) at the Department of Obstetrics and Gynecology, Nihon University Itabashi Hospital between September 2018 and February 2024. Tumor tissue and preoperative plasma samples were available for all cases. This study was approved by the Ethics Review Board of Nihon University Itabashi Hospital (approval number: RK-170711-6) and the University of Tokyo (approval number: G0637- 27), with written informed consent from all participants. HCT116 (ATCC CCL-247; HPV-negative), HeLa (ATCC CCL-2; HPV18-positive), SiHa (ATCC HTB-35; HPV16-positive), and CaSki (ATCC CRL-1550; HPV16-positive) cell lines were used as controls to assess assay specificity.

2. Blood collection and plasma processing

Blood samples were collected in ethylenediaminetetraacetic acid (EDTA) tubes within 30 minutes before surgery, and plasma was separated within 6 hours of blood collection by centrifugation at 1,500×g for 15 minutes. Plasma samples were stored at −80°C until DNA extraction.

3. DNA extraction

DNA was extracted from formalin-fixed paraffin-embedded (FFPE) tumor tissue sections or cell lines using the QIAamp DNA FFPE Advanced UNG Kit (Qiagen, Hilden, Germany) or NucleoSpin DNA FFPE XS (Takara Bio Inc., Shiga, Japan). EDTA plasma was centrifuged at 1,500×g for 15 minutes to remove cellular components, and cell-free DNA (cfDNA) was extracted from 1 mL of plasma using the QIAamp Circulating Nucleic Acid (Qiagen) or Maxwell RSC cfDNA Plasma (Promega, Madison, WI, USA) Kits. DNA concentration was measured using the Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific, Waltham, MA, USA).

4. HPV detection and genotyping by the PGMY method

The PGMY09/11 PCR is a well-established consensus PCR method for broad HPV genotype screening, capable of detecting a wide spectrum of HPV types, including both high-risk and low-risk genotypes, and has been widely used as a reference method in HPV testing, as described in the World Health Organization (WHO) HPV Laboratory Manual and the updated WHO HPV Laboratory e-Manual [18,19]. HPV DNA was amplified using PGMY09/11 consensus primer set [18,19]. PCR was performed in a 50 µL reaction containing 5.0 µL of 10× PCR Buffer II (Thermo Fisher Scientific), 6.0 µL of 25 mM MgCl₂, 5.0 µL of 10 mM dNTP mix, 4.0 µL each of PGMY11 and PGMY09 primer mix, 0.25 µL of AmpliTaq Gold DNA Polymerase (5 U/µL; Thermo Fisher Scientific), and 5.0 µL of extracted DNA. Volume was adjusted with sterile distilled water. Cycling conditions were: initial denaturation at 95°C for 9 minutes; 45 cycles of denaturation at 95°C for 30 seconds, annealing at 55°C for 90 seconds, and extension at 72°C for 2 minutes; followed by final extension at 72°C for 5 minutes. PCR products were stored at 4°C until analysis. Amplification was confirmed by 2% agarose gel electrophoresis, and products were visualized with Midori Green Advance (NE-MG04; Nippon Genetics, Tokyo, Japan) under UV illumination. A 450 bp band indicated HPV positivity. For HPV genotype determination, PCR products were analyzed by reverse line blot hybridization using type-specific oligonucleotide probes on a membrane strip (Table S1). Hybridization and detection were performed per manufacturer protocol, using streptavidin–horseradish peroxidase and chemiluminescence. Genotyping results were interpreted based on the manual.

5. Rationale for dual HPV detection strategy

To comprehensively assess HPV status and viral burden, we employed both PGMY09/11 consensus PCR and genotype-specific dPCR as complementary approaches. PGMY09/11 PCR was used for initial broad-spectrum HPV screening to identify a wide range of HPV genotypes in tumor tissue, whereas dPCR was applied for sensitive and quantitative detection of 6 predefined high-risk HPV types in both tumor tissue and plasma.

6. HPV detection and quantification by dPCR

To achieve higher sensitivity and absolute quantification of HPV DNA, genotype-specific dPCR assays were employed. The nanoplate-based dPCR system enables absolute quantification of target molecules without the need for standard curves and provides enhanced sensitivity for low-copy detection, particularly in samples with limited or fragmented DNA. HPV quantification from cell lines, tumors, and plasma ctDNA was performed using the QIAcuity Digital PCR System (Qiagen), a nanoplatform-based dPCR method. Reaction mixtures containing master mix, probe, primer, and DNA samples were partitioned into nanoplates and subjected to PCR. Absolute quantification was achieved using Poisson statistics based on endpoint amplification. To accommodate the fragmented nature of ctDNA, assays were designed with primers generating amplicons ≤120 bp. Genotype-specific assays targeting HPV16, 18, 31, 39, 45, and 52 were developed based on previously reported methods [20], with primers and probes targeting the E6 and E7 regions of high-risk HPV genotypes (Table S1). Synthetic oligonucleotides for each HPV type (Integrated DNA Technologies, Inc., Coralville, IA, USA) and genomic DNA from HCT116 cell lines for RPP30 were used as positive controls. RPP30 served as an internal control for sample quality and DNA input normalization. The positive control contained 10,000 copies each of HPV and RPP30. The analytical sensitivity (limit of detection [LoD]) of the HPV assays was determined by serial dilution of HPV16/18/31/45/52 oligonucleotides with HCT116 genomic DNA (HPV oligos: HCT116 DNA=1:100, 1:1,000, 1:10,000), tested in triplicate. Detection of the PIK3CA p.E545K mutation was performed using a commercially available assay (dPCR LNA Mutation Assays; Qiagen).

7. Targeted sequencing

The extracted DNA was transferred to DNA CHIP Research Inc. (Kanagawa, Japan) for quality control (QC), library preparation, sequencing, and analysis. Additional QC included DNA integrity assessment by capillary electrophoresis using the Agilent 4200 TapeStation (Agilent Technologies, Santa Clara, CA, USA) and QuantiMIZE DNAQC (Thermo Fisher Scientific) systems. Sequencing libraries were prepared using QIAseq Targeted DNA Pro Comprehensive Cancer Research Panel (PHS-3000Z; Qiagen), which targets 284 cancer-related genes (Table S2), based on the manufacturer’s instructions. Library indices were incorporated using the QIAseq Targeted DNA Pro UDI Kit (Qiagen). Libraries were sequenced on an Illumina NextSeq 500 platform (Illumina, San Diego, CA, USA) with paired-end 150 bp reads, targeting an average coverage depth of approximately 1,500× per locus. Sequencing data were analyzed using the Qiagen Data Analysis Center and custom pipelines at DNA CHIP Research Inc. Reads were aligned to the human reference genome (hg38/GRCh38). Variant calling was performed to identify single-nucleotide variants and small insertions/deletions (indels). Variants were annotated based on their impact on protein function and allele frequency using multiple databases, including 8 Japanese control samples, the Kyoto University HGVD database (Japanese allele frequencies), gnomAD exome data, COSMIC, OncoKB, and other public resources. The impact of each variant was evaluated based on functional predictions and population frequency.

8. Statistical analysis

All statistical analyses were performed using GraphPad Prism (version 10.31; GraphPad Software, San Diego, CA, USA). Continuous variables were analyzed with the Mann-Whitney U test. Categorical variables were compared using the χ² or Fisher’s exact test as appropriate. Correlations between ctHPV DNA levels and clinicopathological factors were assessed using Spearman’s rank correlation coefficient. For univariable analyses, the association between ctHPV DNA copy number and each clinicopathological factor was evaluated individually. Variables showing statistical significance in univariable analyses were subsequently included in multivariable linear regression analysis to assess independent associations with ctHPV DNA copy number. Given the limited sample size, a parsimonious model was adopted to avoid overfitting. A p-value of <0.05 was considered statistically significant.

RESULTS

1. Patient characteristics

We analyzed data from 49 patients with cervical cancer (Table 1). The median age was 44 years (range, 28–80). Based on the 2018 FIGO classification, 31 patients (63.3%) were stage I, 5 (10.2%) stage II, and 13 (26.5%) stage III (T1 or T2/N1). By TNM criteria, 36 (73.5%) and 13 (26.5%) patients had T1 and T2 tumors, respectively. The median tumor size was 35 mm (range, 5.5–70 mm). All procedures were performed via an abdominal approach, including radical hysterectomy in 42 patients (85.7%) and radical trachelectomy in 5 (10.2%). Histology revealed 32 (65.3%) SCC, 14 (28.6%) adenocarcinoma, and 3 (6.1%) other types. Lymph node metastasis was present in 12 patients (24.5%), and lymphovascular invasion was positive in 26 patients (53.1%, Table 1).

Table 1. Patient characteristics of 49 cervical cancer.

Characteristics		Value
Age (yr)		44 (28–80)
FIGO
	IA1	1 (2.0)
	IB1	13 (26.5)
	IB2	14 (28.6)
	IB3	3 (6.1)
	II	5 (10.2)
	III (T1or2/N1)	13 (26.5)
Tumor size (mm)		35 (5.5–70)
T stage
	T1	36 (73.5)
	T2	13 (26.5)
Surgical procedure
	Radical hysterectomy	42 (85.7)
	Modified radical hysterectomy	1 (2.0)
	Radical trachelectomy	5 (10.2)
	Simple hysterectomy	1 (2.0)
Histology
	Squamous cell carcinoma	32 (65.3)
	Adenocarcinoma	14 (28.6)
	Others	3 (6.1)
Lymph node metastasis
	Positive	12 (24.5)
	Negative	37 (75.5)
Lymphovascular invasion
	Positive	26 (53.1)
	Negative	23 (46.9)
Surgical margin status
	Positive	6 (12.2)
	Negative	43 (87.8)
Parametrial invasion
	Positive	5 (10.2)
	Negative	44 (89.8)
Depth of stromal invasion (mm)
	≤5	8 (16.3)
	>5	40 (81.6)
	Unknown	1 (2.0)
Tumor diameter (cm)
	≥4	20 (40.8)
	<4	29 (59.2)

Values are presented as number (%) or median (range). All procedures were performed via an abdominal approach.

FIGO, International Federation of Gynecology and Obstetrics.

2. Analytical sensitivity and HPV type specificity of the dPCR assays

The LoD for the genotype-specific HPV dPCR assays was assessed using serial 10-fold dilutions of HPV16/18/31/45/52 reference DNA (oligos) in a background of HPV-negative human genomic DNA from HCT116 cells. Each assay consistently detected 0.01% target DNA (1 copy/10,000) with no false positives; LoD data are in Table 2 and representative HPV39 results in Fig. 1A and B. Analytical sensitivity was assessed using HPV16-positive SiHa and CaSki, and HPV18-positive HeLa cells (Fig. 1C). HPV16 was detected at 3.89 copies/cell in SiHa and 1,048.8 copies/cell in CaSki, and HPV18 at 7.25 copies/cell in HeLa, with no cross-detection (Fig. 1D). Our results were consistent with prior reports, which describe 10–50 copies of HPV18 in HeLa, 1–6.5 copies of HPV16 in SiHa, and 400–600 copies of HPV16 in CaSki [21,22].

Table 2. Type-specific HPV detection sensitivity by digital polymerase chain reaction.

HPV	Dilution	HPV (copies/μL)	Endogenous control gene (RPP30; copies/μL)	Control (water; copies/μL)	Input HPV (copies/40 μL)	Input (RPP30; copies/40 μL)	Total input (copies/40 μL)
HPV16	1:100	3.870	977	0	154.8	39,092	39,247
	1:1,000	0.855	894	0	34.2	35,756	35,790
	1:10,000	0.215	1,189	0	8.6	47,548	47,557
HPV18	1:100	26.358	206	0	1054.3	8,240	9,294
	1:1,000	0.329	133	0	13.2	5,310	5,323
	1:10,000	0.383	820	0	15.3	32,816	32,831
HPV31	1:100	3.963	316	0	158.5	12,628	12,787
	1:1,000	1.091	106	0	43.6	4,244	4,288
	1:10,000	0.109	986	0	4.4	39,440	39,444
HPV39	1:100	7.347	1,015	0	293.9	40,588	40,882
	1:1,000	1.276	982	0	51.0	39,260	39,311
	1:10,000	0.162	1,016	0	6.5	40,640	40,646
HPV45	1:100	12.290	1,011	0	491.6	40,448	40,940
	1:1,000	2.235	989	0	89.4	39,576	39,665
	1:10,000	3.972	1,030	0	158.9	41,208	41,367
HPV52	1:100	7.150	580	0	286.0	23,204	23,490
	1:1,000	1.944	125	0	77.8	4,996	5,073
	1:10,000	0.219	861	0	8.8	34,428	34,437

HPV, human papillomavirus.

Fig. 1. Sensitivity and specificity of dPCR assays for HPV genotyping. (A) Representative dPCR plots for HPV39 using 10-fold serial dilutions of plasmid DNA (0.01%–1%). RPP30 served as an internal control; no amplification was observed in no-template controls. (B) Quantification of HPV39 dilution series by dPCR. (C) Assay specificity. Fluorescence intensity for the HPV target is displayed on the y-axis and fluorescence intensity for RPP30, used for copy number normalization, on the x-axis, enabling clear discrimination of double-positive, single-positive, and negative partitions. HPV16 was detected exclusively in SiHa and CaSki cells (HPV16-positive) with no HPV18 detection, whereas HPV18 was detected exclusively in HeLa cells (HPV18-positive) with no HPV16 detection, confirming the assay’s specificity. (D) The copy number per cell measured in cell lines. (E) Two-dimensional dPCR scatter plots of FFPE samples analyzed by dPCR for 6 high-risk HPV genotypes (HPV16, 18, 31, 39, 45, and 52). (F) Concordance of type-specific HPV detection between PGMY and dPCR results in the initial 24 FFPE tumor samples.

dPCR, digital polymerase chain reaction; FFPE, formalin-fixed paraffin-embedded; HPV, human papillomavirus; RFU, relative fluorescence units.

3. HPV genotyping in FFPE tumor samples

FFPE tumor samples were genotyped using both the PGMY method and the dPCR-based assay. In the initial 24 cases, the PGMY method identified 31 HPV genotypes (Fig. S1). To verify the performance of the genotype-specific dPCR assays, we first tested representative FFPE tumor DNA for 6 high-risk HPV types (HPV16, 18, 31, 39, 45, and 52), confirming clear amplification signals (Fig. 1E). Owing to limited sample availability, we focused on 4 representative genotypes (HPV16, 18, 31, and 52) and analyzed the initial 24 cases by dPCR for comparison with the PGMY data.

Using the PGMY method, HPV DNA was detected in 19 cases (79.2%), comprising HPV16 (11), HPV18 (2), HPV31 (1), HPV33 (2), HPV45 (1), and HPV52 (1). However, dPCR identified additional HPV types, detecting HPV in 4 of the 5 cases that were PGMY-negative. All genotypes identified by PGMY were also detected by dPCR. In total, HPV DNA was detected in 23 of 24 cases (95.8%), including the single HPV45 case detected only by PGMY (Table S3). Single-genotype infection was observed in 8 cases (33.3%), whereas multiple-genotype infections occurred in 15 cases (62.5%), with concurrent HPV16 and HPV18 infections in 9 cases (37.5%). Overall, the combined genotype distribution was HPV16 in 19 cases, HPV18 in 12, HPV31 in 2, HPV33 in 2, HPV45 in 1, and HPV52 in 5. Notably, dPCR detected more HPV18 and HPV52 cases than PGMY, demonstrating its superior sensitivity in FFPE tumor samples (Fig. 1F).

Subsequently, in the remaining 25 cases, we evaluated concordance between dPCR-based genotyping and high-risk HPV genotypes (HPV16, 18, 31, 39, 45, and 52) identified using the PGMY method. High-risk HPV genotypes were detected using the PGMY method in 19 of 25 cases (76.0%), and in all 19 cases (100%), the corresponding genotype was confirmed by dPCR (Table S4). Across all 49 tumor samples analyzed for HPV genotypes using either the PGMY method, dPCR, or both, HPV genotypes were identified in 42 cases (85.7%). The genotype distribution in these 42 cases was HPV16 in 26 cases (61.9%), HPV18 in 6 (14.3%), and other types (HPV31, 39, 45, and 52) in 10 (23.8%) (Fig. 2A).

Fig. 2. ctHPV DNA detection and association with clinicopathological features in surgically resected cervical cancer. (A) HPV genotype distribution in tumor tissues (n=42). (B) Detection rate of ctHPV DNA in preoperative plasma stratified by HPV genotype. (C) Maximum tumor diameter and ctHPV DNA. (D-G) ctHPV DNA copy numbers by T stage (D), histology (E), lymphovascular invasion (F), and lymph node metastasis (G). (H) Multivariable linear regression analysis of lymphovascular invasion and lymph node metastasis in relation to ctHPV DNA copy number (n=41). One patient with missing lymph node status was excluded. Regression coefficients with 95% confidence intervals are shown.

ctHPV, circulating tumor human papillomavirus; HPV, human papillomavirus; SCC, squamous cell carcinoma.

^*p<0.05.

4. Detection and quantification of ctHPV DNA in plasma and comparison with tumor tissue HPV DNA

As a reference, we first quantified HPV DNA copy numbers in 42 FFPE tumor samples by dPCR (Table 3). HPV DNA was quantifiable in 41 of 42 tumor tissues (97.6%), with a median copy number of 8.4 per reaction (range, 0–705.2; mean ± standard deviation [SD], 51.3±117.8). Notably, 7 cases (16.7%) yielded <1 copy per reaction, including 4 cases (9.5%) with <0.1 copy, indicating that although most FFPE samples contained detectable HPV DNA, the quantitative levels varied widely across cases, likely reflecting differences in tumor cellularity, DNA integrity, or intrinsic HPV copy number.

Table 3. Summary of detection and quantification of type-specific HPV DNA with FFPE and liquid biopsy.

Case No.	HPV typing by tissue FFPE	Quantification by tissue FFPE		Detection and quantification by ctDNA
		HPV†	Control (RPP30)‡	HPV§	Control (RPP30)‡
1	16	254.49	98.4	0	10.94
2	16	21.77	81.5	0	8.17
3	16*,18*,52	37.68	112.1	595.6	86.49
4	16,52*	208.14	161.9	0	191.10
5	18*,52*	15.64	86.9	22.0	42.03
6	16*,18*,33,52*	0.06	31.1	0	20.00
7	16,18*	133.40	20.0	0	10.92
8	16*,31,52*	0.63	123.0	21.2	82.51
9	16,18*	2.82	51.1	0	84.31
10	16	705.23	51.1	21.2	37.51
11	45	16.33	4,163.2	0	34.89
12	16	33.15	23.1	0	9.92
13	18*	0	315.6	0	11.38
14	16,18*	133.37	48.1	144.8	9.00
15	16*,18	47.21	126.0	0	19.35
16	16,18*	29.05	156.6	120.4	17.34
17	18	47.88	111.5	0	14.25
18	Neg	NA	NA	NA	NA
19	16*	4.17	31.8	0	50.36
20	16,18*	52.54	107.6	0	56.04
21	16,31*	81.08	120.4	106.4	5.33
22	16	0.16	110.6	0	96.73
23	16*,18*	0.09	104.2	0	36.63
24	16*,33	0.01	100.2	0	12.99
25	16	4.28	393.3	0	0.16
26	Neg	NA	NA	NA	NA
27	Neg	NA	NA	NA	NA
28	Neg	NA	NA	NA	NA
29	18	8.07	1,041.5	1,050.4	0.59
30	16	2.90	93.6	0	0.00
31	45	8.37	1,176.2	0	0.00
32	39	8.14	1,405.9	0	0.05
33	31	4.41	528.5	0	6.84
34	16	66.20	2,409.5	0	4.46
35	45	10.25	6,720.7	0	112.70
36	16	47.87	786.8	21.2	5.13
37	16	107.96	4,059.8	236.4	5.86
38	Neg	NA	NA	NA	NA
39	Neg	NA	NA	NA	NA
40	16	5.15	766.8	0	6.14
41	16	1.59	58.7	22.0	0.11
42	16	1.85	121.2	62.8	6.42
43	39	2.60	292.5	0	0.00
44	16	2.70	216.8	0	11.40
45	16	8.42	608.7	0	0.70
46	52	35.58	599.7	0	0.05
47	Neg	NA	NA	NA	NA
48	16	2.28	123.5	0	0.05
49	18	0.88	57.2	0	0.00

HPV typing was performed on tissue FFPE samples, and the corresponding HPV DNA copy numbers were quantified from both FFPE and plasma (ctDNA). The second column indicates HPV genotypes detected by FFPE.

ctDNA, circulating tumor DNA; FFPE, formalin-fixed paraffin-embedded; HPV, human papillomavirus; NA, not available; Neg, negative.

^*Asterisks denote HPV genotypes identified by FFPE but not subjected to copy number quantification.

^†Values are expressed as copies per tumor cell; ^‡Values as copies per μL; ^§Values as copies per 1 mL of plasma.

Subsequently, we examined whether corresponding ctHPV DNA could be detected in the plasma. The median cfDNA yield was 17.2 ng (interquartile range: 11.58–34.70 ng, range: 6.80–98.80 ng, n=42), confirming adequate sample quality for downstream analyses. Among 42 cases tested, ctHPV DNA was identified in 12 cases, corresponding to a detection rate of 28.6%. Copy numbers varied considerably, ranging from 21.2 to 1,050 copies per 1 mL of plasma (median of 0 copies/mL; mean ± SD 57.7±187) (Table 3). Detection rates by genotype were 30.8% for HPV16 (8/26), 33.3% for HPV18 (2/6), and 20% for other types (Fig. 2B), with no significant differences among genotypes.

5. Association between ctHPV DNA copy number and clinical parameters

We investigated the association between ctHPV DNA copy number and clinical parameters, including maximum tumor diameter, T stage, and histological types. With respect to tumor size, no significant association was observed between maximal tumor diameter and ctHPV DNA copy number. Specifically, Spearman rank correlation analysis showed no correlation between tumor size and ctHPV DNA copy number (r=0.021, p=0.893). In addition, when tumor size was analyzed as a categorical variable (<4 cm vs. ≥4 cm), ctHPV DNA copy numbers did not differ significantly between the 2 groups (p=0.962) (Fig. 2C). Similarly, ctHPV DNA levels did not differ significantly between T1 (n=33) and T2 (n=9) tumors (Fig. 2D), between SCC (n=27) and non-SCC histologies (n=15) (Fig. 2E). All the 4 recurrent cases were negative for ctHPV DNA. Univariable analyses were additionally performed to evaluate the associations between ctHPV DNA copy number and a comprehensive set of clinicopathological parameters, including surgical margin status, parametrial invasion, depth of cervical stromal invasion, age, and histology. The detailed results of these analyses are summarized in Table S5. These results indicate that the ctHPV DNA copy number was not associated with these clinicopathological characteristics in this cohort.

Subsequently, we examined other pathological features, i.e., lymphovascular invasion and lymph node metastasis (Fig. 2F and G). ctHPV DNA copy number was significantly higher in cases with positive lymphovascular invasion than in those without (p=0.034, Fig. 2F). Moreover, ctHPV DNA copy number was also significantly higher in cases with lymph node metastasis (p=0.010, Fig. 2G). These findings suggest that ctHPV DNA levels may be associated with lymphovascular invasion and lymph node metastasis.

To assess whether these associations were independent, we further performed multivariable linear regression analysis including lymphovascular invasion and lymph node metastasis. In this model, lymph node metastasis remained significantly associated with ctHPV DNA copy number, whereas lymphovascular invasion did not (Fig. 2H). These results indicate that lymph node metastasis is an independent pathological factor associated with elevated ctHPV DNA levels in surgically resected cervical cancer.

6. Lack of plasma ctDNA detection of PIK3CA hotspot mutations found by tumor panel sequencing highlights the sensitivity of ctHPV DNA

Targeted sequencing was performed on tumor tissues from 9 surgically resected patients to identify driver mutations and compare their detectability in ctDNA with that of ctHPV DNA. Targeted sequencing of tumor tissues from 9 surgically resected patients identified 24 mutations in 7 cases: 15 missense, 6 nonsense/stop-gain, and 3 frameshift. Among them, 14 variants were annotated as pathogenic or likely pathogenic, including PIK3CA (4 cases; p.E545K in 3 cases), KMT2D (3), FBXW7 (2), CREBBP (2), EP300 (1), and ARID1A (1) (Table 4). The recurrent hotspot PIK3CA p.E545K was one of the well-known hotspots in this gene.

Table 4. Summary of driver mutations identified by tumor tissue comprehensive genomic profiling in 9 cases.

Case No.	Gene	Variant	VAF (%)	Pathogenicity according to ClinVar*/OncoKB† databases	Transcript (RefSeq)	ClinVar ID	COSMIC ID	Plasma Sample_dPCR
1	PIK3CA	c.1633G>A, p.Glu545Lys	8.4	Pathogenic*	NM_006218.4	VCV000013655.61	COSV55873239;13655	No data‡
2	PIK3CA	c.1258T>C, p.Cys420Arg	21.0	Pathogenic*	NM_006218.4	VCV000031945.51	COSV55874020;31945	No data‡
	ARID1A	c.631C>T, p.Gln211Ter	13.0	Likely pathogenic† (stopgain)	NM_006015.6			No data‡
	FBXW7	c.73_74delCT, p.Leu25fs	2.7	Likely pathogenic† (out of frame)	NM_001349798.2			No data‡
	FBXW7	c.1531G>A, p.Gly511Ser	1.6	Unknown oncogenic effect†	NM_001349798.2		COSV9966263	No data‡
	KMT2D	c.15700G>T, p.Glu5234Ter	4.0	Likely pathogenic† (stopgain)	NM_003482.4		COSV56503806	No data‡
3	KMT2D	c.6177delC, p.Tyr2060fs	3.8	Likely pathogenic† (stopgain)	NM_003482.4			No data‡
	FBXW7	c.1433C>T, p.Ser478Phe	2.5	Likely pathogenic*	NM_001349798.2	VCV003093874.1	COSV55954331	No data‡
	PIK3CA	c.2524G>A, p.Gly842Ser	3.0	Uncertain significance*	NM_006218.4	VCV000565987.8	COSV104566914;565987	No data‡
4	PIK3CA	c.1633G>A, p.Glu545Lys	14.3	Pathogenic*	NM_006218.4	VCV000013655.61	COSV55873239;13655	Not detected
	CREBBP	c.2515C>T, p.Gln839Ter	1.2	Likely pathogenic† (stopgain)	NM_004380.3		COSV52120786	No data‡
	EP300	c.4564C>T, p.Gln1522Ter	2.4	Likely pathogenic† (stopgain)	NM_001429.4			No data‡
	KMT2D	c.7309C>T, p.Pro2437Ser	1.5	Unknown oncogenic effect†	NM_003482.4		COSV56498539	No data‡
5	FBXW7	c.2033C>A, p.Ser678Ter	1.3	Likely pathogenic† (stopgain)	NM_001349798.2			No data‡
6	CREBBP	c.2224C>T, p.Arg742Cys	6.0	Uncertain significance*	NM_004380.3	VCV002902206.3	COSV52115701	No data‡
	CREBBP	c.2467C>T, p.Gln823Ter	2.7	Likely pathogenic† (stopgain)	NM_004380.3			No data‡
	EP300	c.4529C>G, p.Pro1510Arg	2.9	Uncertain significance*	NM_001429.4	VCV000575459.10		No data‡
	EP300	c.2680C>T, p.Pro894Ser	1.6	Unknown oncogenic effect†	NM_001429.4		COSV54334908	No data‡
	KMT2D	c.12715delC, p.Gln4239fs	4.1	Likely pathogenic† (stopgain)	NM_003482.4			No data‡
7	PIK3CA	c.1633G>A, p.Glu545Lys	14.1	Pathogenic*	NM_006218.4	VCV000013655.61	COSV55873239;13655	Not detected
	PIK3CA	c.2915C>T, p.Thr972Ile	2.2	Uncertain significance*	NM_006218.4	VCV000456544.9	COSV55933090;456544	No data‡
	EP300	c.2109G>A, p.Met703Ile	1.6	Unknown oncogenic effect†	NM_001429.4		COSV54343659	No data‡
	EP300	c.4388T>C, p.Leu1463Pro	5.4	Unknown oncogenic effect†	NM_001429.4		COSV99607535	No data‡
	RB1	c.785G>A, p.Arg262Gln	2.8	Uncertain significance*	NM_000321.3	VCV000842202.11	COSV57304729	No data‡
8	(-)
9	(-)

Among the 9 cases in which tumor tissue was submitted for next generation sequencing, driver gene mutations were identified in 7 cases. No PIK3CA mutations detected in ctDNA, despite their presence in surgical specimens. #CREBBP p.Gln823^* is considered likely oncogenic, considering the pathogenicity of p.Q1113^* and other stopgain mutations. Cases 4 and 7: ctDNA tested for the hotspot PIK3CA p.E545K mutation.

ctDNA, circulating tumor DNA; VAF, variant allele frequency.

^*Annotated using ClinVar; ^†Annotated using oncoKB; ^‡Not performed.

Subsequently, we evaluated whether the ctDNA assay could detect the PIK3CA p.E545K variant in 2 cases with available plasma using dPCR. The mutation was not detected in either case, indicating that ctDNA harboring PIK3CA mutations was below the assay detection limit. Conversely, ctHPV DNA was detectable in both cases. Collectively, these observations suggest that the detection of hotspot driver mutations in ctDNA may be more limited and potentially less sensitive than that of ctHPV DNA.

DISCUSSION

In this study, we assessed ctHPV DNA in preoperative plasma from patients with surgically resected cervical cancer using dPCR. ctHPV DNA was detected in 28.6% of cases, and the HPV genotypes identified in plasma were concordant with those in the corresponding tumor tissues. Notably, ctHPV DNA copy numbers were significantly associated with lymphovascular invasion and lymph node metastasis, suggesting its potential as a noninvasive biomarker of high-risk pathological features in surgically resected cervical cancer.

The ctHPV DNA detection rate in our cohort was lower than that reported in some recent studies of advanced cervical cancer [16,17]. However, this finding is consistent with reports focusing on early-stage cervical cancer, in which tumor burden is limited [15]. Given the adequate analytical sensitivity of our dPCR assay and sufficient cfDNA yield, the observed detection rate is more likely to reflect the biological characteristics of surgically resected disease rather than primarily methodological limitations.

ctDNA levels often reflect systemic disease progression, particularly hematogenous dissemination [23,24], but may also indicate lymphatic spread when associated with lymphovascular invasion or increased tumor burden [23,25]. Our findings suggest that ctHPV DNA may indicate early tumor cell dissemination, including lymphatic involvement, even in localized cervical cancer. Notably, ctDNA enables the early detection of regional lymph node relapse in lung cancer [26], and in HPV-related oropharyngeal cancer, post-treatment ctHPV DNA predicts nodal recurrence [27]. These observations reinforce the relevance of ctHPV DNA in capturing the early metastatic potential of cervical cancer.

Our findings confirm the superior sensitivity of dPCR over conventional PGMY09/11 consensus PCR. While PGMY targets ~450-bp amplicons and often fails with fragmented DNA or low viral loads [28,29,30,31], our dPCR assays utilize short amplicons (≤120 bp) and DNA partitioning, enabling robust detection even in FFPE and plasma [15]. This enhanced sensitivity likely explains the higher rate of multiple-genotype infections detected (63%) compared to previous reports (<30%) [32], highlighting dPCR’s utility for samples with limited viral burden.

While ctHPV DNA is established in advanced cervical cancer, its utility in early-stage disease remains under investigation. In our cohort, ctHPV DNA correlated with lymphovascular invasion and lymph node metastasis but not with tumor size or stage, suggesting that it reflects subclinical tumor aggressiveness rather than tumor bulk alone. Given that current imaging and serum markers often fail to detect microscopic nodal disease [33,34], ctHPV DNA may serve as a useful noninvasive biomarker to refine preoperative risk stratification and complement standard surgical assessments [27,35].

Recent studies have further supported the clinical value of ctHPV DNA in cervical cancer. Notably, ctHPV DNA was detected in 31.6% of early-stage and 90% of advanced-stage cases, with no positivity in non-malignant controls, confirming its high specificity [36]. In a prospective multicenter cohort, persistent ctHPV DNA after chemoradiation predicted poor progression-free survival [37]. These findings emphasize the stage-stratified prognostic value of ctHPV DNA and its potential integration into individualized treatment strategies [17]. In this context, ctHPV DNA testing can be incorporated into preoperative assessment algorithms. Rather than serving as a standalone decision threshold, preoperative ctHPV DNA positivity may function as a complementary risk indicator reflecting early lymphatic dissemination. For example, patients with detectable ctHPV DNA may warrant more careful evaluation during sentinel lymph node assessment and more thorough preoperative counseling when considering fertility-sparing surgery, given the potential for lymph node involvement. With further prospective validation, including evaluation of correlations with pathological features, and recurrence patterns, such information could support individualized surgical planning and risk stratification.

Previous studies have also demonstrated the clinical utility of ctHPV DNA in monitoring treatment response and detecting recurrence in HPV-associated malignancies, supporting its broader role in therapeutic decision-making, even in curative settings [27]. A major advantage of liquid biopsy is its ability to detect minimal residual disease (MRD) and systemic dissemination beyond the resolution of standard imaging [24]. Serial quantitative measurement of ctHPV DNA after surgery, as a form of MRD assessment, could guide decisions regarding adjuvant radiotherapy in patients with lymphovascular space invasion [37,38].

In addition, ctHPV DNA offers a technical advantage over hotspot mutations: HPV genomes are tumor-specific and often multi-copy, yielding higher copy numbers and variant fractions detectable by dPCR, whereas mutant ctDNA is diluted by abundant wild-type DNA. In this context, PIK3CA was examined as a representative example of a conventional mutation-based ctDNA approach. The lack of detectable PIK3CA-mutated ctDNA under identical analytical conditions further supports the practical advantage of ctHPV DNA as a liquid biopsy biomarker. Overall, these findings suggest that ctHPV DNA could improve risk stratification and support individualized surgical planning in cervical cancer.

This study had some limitations. First, it was a single-institution analysis with a modest sample size, and the limited number of ctHPV DNA–positive cases reduced the statistical power for subgroup analyses, including genotype-specific comparisons. Second, the retrospective design and use of archived plasma samples introduced variability in sample volume and quality, which may have influenced detection sensitivity. Third, as our genotype-specific dPCR assays were limited to 6 high-risk HPV types (HPV16, 18, 31, 39, 45, and 52), ctHPV DNA derived from other regionally prevalent genotypes, such as HPV33 and HPV58, may not have been detected. However, single-type infections with these genotypes are uncommon in Japan (approximately 2%–4% for HPV33 and 3%–6% for HPV58) [39,40], and analysis of tumor tissues in a subset of 24 cases (Table S3) revealed that HPV33 occurred only as part of multiple infections with genotypes included in our panel, and HPV58 was not detected. These findings suggest that the impact of this limitation on ctHPV DNA detection was likely minimal in this cohort. To address these limitations, future prospective multicenter studies with standardized plasma collection, larger cohorts and expanded genotype panels are needed to validate genotype-specific differences, establish ctHPV DNA as a predictive biomarker for lymphovascular invasion and lymph node metastasis, and further explore its potential role in recurrence monitoring.

SUPPLEMENTARY MATERIALS

Table S1

Primers for type-specific, high-risk HPV, designed in the E6 and E7 regions

Table S2

List of 284 genes targeted in the QIAseq targeted DNA pro panel

Table S3

HPV genotyping of 24 FFPE tumor samples by either PGMY or dPCR method

Table S4

HPV genotyping of additional 25 FFPE tumor samples by either PGMY or dPCR method

Table S5

Univariable analysis of clinical factors associated with circulating tumor human papillomavirus DNA copy number

Fig. S1

HPV genotypes covered by the PGMY assay.