Blood-based screening for HPV-associated cancers

Abstract

Background: HPV-associated oropharyngeal cancer (HPV+OPSCC) is the most common HPV-associated cancer in the United States yet unlike cervical cancer lacks a screening test. HPV+OPSCCs are presumed to start developing 10-15 years prior to clinical diagnosis. Circulating tumor HPV DNA (ctHPVDNA) is a sensitive and specific biomarker for HPV+OPSCC. Taken together, blood-based screening for HPV+OPSCC may be feasible years prior to diagnosis.

Methods: We developed an HPV whole genome sequencing assay, HPV-DeepSeek, with 99% sensitivity and specificity at clinical diagnosis. 28 plasma samples from HPV+OPSCC patients collected 1.3-10.8 years prior to diagnosis along with 1:1 age and gender-matched controls were run on HPV-DeepSeek and an HPV serology assay.

Results: 22/28 (79%) of cases and 0/28 controls screened positive for HPV+OPSCC with 100% detection within four years of diagnosis and a maximum lead time of 7.8 years. We next applied a machine learning model classifying 27/28 cases (96%) with 100% detection within 10 years. Plasma-based PIK3CA gene mutations, viral genome integration events and HPV serology were used to orthogonally validate cancer detection with 68% (19/28) of the cohort having multiple cancer signals detected. Molecular fingerprinting of HPV genomes was performed across patients demonstrating that each viral genome was unique, ruling out contamination. In patients with tumor blocks from diagnosis (15/28), molecular fingerprinting was performed within patients confirming the same viral genome across time.

Conclusions: We demonstrate accurate blood-based detection of HPV-associated cancers with lead times up to 10 years before clinical cancer diagnosis and in doing so, highlight the enormous potential of ctDNA-based cancer screening.

Figure 1.. HPV-DeepSeek workflow, training and diagnostic performance A.

Schematic representation of HPV-DeepSeek workflow. B. Determination of HPV-DeepSeek threshold using serial dilutions of HPV+OPSCC patient cfDNA into control cfDNA, performed in triplicate. Circle size represents the quantity of HPV+OPSCC patient cfDNA input. Threshold was set at 10 unique HPV genome reads and 10% HPV genome coverage. Samples below the test threshold and considered negative are in the gray shaded region. White shaded region contains positive samples. The x-axis represents the number of HPV reads and the y-axis represents HPV genome coverage as percent of total genome. C. HPV-DeepSeek diagnostic performance in 153 HPV+OPSCC cases and 153 population-level controls applying the pre-determined thresholds. HPV+OPSCC cases are green, controls are red. HPV genotypes are represented in different shapes. Sensitivity 98.7%, Specificity 98.7%. D. Receiver operating curves for three machine learning models for classifying HPV+OPSCC vs. no HPV+OPSCC. The AUC values are shown for all three models based on risk scores on the test cohort with 95% confidence intervals (blue, RandomForest; green, AdaBoost; red, NaiveBayes).

Figure 2.. Screening detection of HPV+OPSCC A.

Histogram demonstrating blood sample collection timepoint in years prior to diagnosis for 28 HPV+OPSCC patients. B. Histogram showing results of HPV-DeepSeek in 28 HPV+OPSCC patient samples demonstrating 22/28 (79%) samples screening positive. Samples in green are positive for HPV+OPSCC based on the pre-set cutoffs. Samples in red are negative. C. HPV-DeepSeek results of 56 samples (28 HPV+OPSCC, green color; 28 age- and gender- matched controls, red color). The colors of green correspond to the time of collection, with light green representing samples collected closer to diagnosis and the dark green representing samples collected further from diagnosis. Lighter-green samples tend to have higher numbers of reads and higher coverage and darker-green samples tend to have lower numbers of reads and lower coverage supporting increasing ease of detection closer to time of diagnosis. D. Histogram showing results of HPV-DeepSeek with machine learning demonstrating 27/28 (96%) samples screening positive. E. Bar graph showing the comparison between HPV-DeepSeek and HPV-DeepSeek with machine learning. Data is divided in tertiles based on collection time point. The deep green color represents the percentage of cases determined positive by HPV-DeepSeek. The light green color represents the cases detected with the addition of machine learning, showing improved performance with machine learning for samples further from the time of diagnosis.

Figure 3.. Orthogonal validation of cancer detection in the blood A.

Histogram representing the 28 HPV+OPSCC samples ordered by sample collection time in years prior to diagnosis. Below, features supportive of cancer detection. Samples positive for a given feature are in green, samples negative for a feature are in red, and untested samples are in beige. B. HPV-human genome integration events detected in seven patients. C. Venn diagram for HPV serology and ctHPVDNA results showing 20/26 patients had HPV oncoprotein antibodies detected of which 17 also had ctHPVDNA detected. D. Bar plot indicating the number of samples with 0, 1, 2 or 3 cancer diagnosis-supporting features present. 25/28 cases had at least one cancer diagnosis supporting feature and 19/28 had at least two features.

Figure 4.. Viral genome molecular fingerprinting and longitudinal cancer monitoring A.

Phylogenetic tree of the 28 HPV+OPSCC samples based on the genotype, lineage and sublineage detected in plasma (left) and molecular fingerprinting heatmap (right) for samples sharing the same HPV16 sublineage. SNVs for each sample are shown in red, demonstrating that each virus is unique. B. Pairwise correlation heatmap for the 15 samples that had FFPE tumor tissue blocks from diagnosis available, showing that plasma-tissue pairs correlate within a pair and not across pairs. The highest correlation of 1 is represented in red while no correlation or 0 is represented in white. C. Longitudinal monitoring of four HPV+OPSCC patients from screening time point to diagnosis, treatment and post-treatment monitoring. Vertical dotted line represents time of diagnosis. Patients 8 and 21 were treated with surgery. Patients 1 and 3 were treated with chemoradiotherapy. In patients 1, 8, and 21 ctHPVDNA is cleared following treatment and remained zero during monitoring. In patient 3, ctHPVDNA was detected 20 months before diagnosis. ctHPVDNA levels were monitored weekly during chemoradiotherapy treatment with decreasing levels, but no clearance. Following conclusion of treatment, ctHPVDNA levels began increasing. The patient was then found to have a second primary HPV malignancy (asterix) for which they underwent surgery followed by chemoradiotherapy. ctHPVDNA cleared after this treatment but re-elevated, indicating recurrence, which was detected by cross-sectional imaging two months later.