ctDNA transiting into urine is ultrashort and facilitates noninvasive liquid biopsy of HPV+ oropharyngeal cancer

Abstract

BACKGROUNDTransrenal cell-free tumor DNA (TR-ctDNA), which transits from the bloodstream into urine, has the potential to enable noninvasive cancer detection for a wide variety of nonurologic cancer types.MethodsUsing whole-genome sequencing, we discovered that urine TR-ctDNA fragments across multiple cancer types are predominantly ultrashort (<50 bp) and, therefore, likely to be missed by conventional ctDNA assays. We developed an ultrashort droplet digital PCR assay to detect TR-ctDNA originating from HPV-associated oropharyngeal squamous cell carcinoma (HPV+ OPSCC) and confirmed that assaying ultrashort DNA is critical for sensitive cancer detection from urine samples.ResultsTR-ctDNA was concordant with plasma ctDNA for cancer detection in patients with HPV+ OPSCC. As proof of concept for using urine TR-ctDNA for posttreatment surveillance, in a small longitudinal case series, TR-ctDNA showed promise for noninvasive detection of recurrence of HPV+ OPSCC.ConclusionOur data indicate that focusing on ultrashort fragments of TR-ctDNA will be important for realizing the full potential of urine-based cancer diagnostics. This has implications for urine-based detection of a wide variety of cancer types and for facilitating access to care through at-home specimen collections.FundingNIH grants R33 CA229023, R21 CA225493; NIH/National Cancer Institute grants U01 CA183848, R01 CA184153, and P30CA046592; American Cancer Society RSG-18-062-01-TBG; American Cancer Society Mission Boost grant MBGI-22-056-01-MBG; and the A. Alfred Taubman Medical Research Institute.

Keywords: Cancer; Diagnostics; Head and neck cancer; Oncology.

Figure 1. Fragment size analysis of urine cfDNA WGS from patients with nonurologic cancers, showing TR-ctDNA enrichment in ultrashort fragments.

(A) The plots show log₂(copy number ratio) (Log2CNRatio) calls on the y axis for plasma cfDNA or urine cfDNA for 3 patients, with the x axis showing the genomic position of the mapped DNA fragments across the indicated chromosomes. Low-pass coverage urine cfDNA WGS data, unfiltered for fragment length, showing CNA patterns qualitatively concordant with those from matched plasma cfDNA WGS (also unfiltered for length) in some but not all patients; cancer-associated CNAs are visible in patients AML14 and AML13 but not in patient ST5. After stratification of analysis by fragment length into 20 bp wide bins, CNA plots showed that restriction to ultrashort bins (<50 bp; i.e., 30–50 bp and 20–40 bp) revealed tumor-associated CNA more robustly than unfiltered data. (B) Heatmap of estimated percentage of tumor DNA content from WGS data. We observed increased enrichment for tumor DNA with ultrashort fragments (i.e., <50 bp) relative to larger fragment size bins, consistent with the qualitative differences evident in A. (C) Composite curves of urine cfDNA WGS data from 3 patients with solid tumors and 3 patients with AML. The curves labeled as focally amplified, gain, and loss are implicitly derived from tumor-enriched DNA (fragment length of mapped reads emanating from regions with known tumor-associated CNA), while the unaltered curves reflect inferred fragment lengths for reads mapping to the remainder of the genome. As shown, the majority of DNA fragments and, importantly, those mapping to genomic regions of tumor-associated CNA are ultrashort (<50 bp) in length.

Figure 2. Ultrashort amplicon HPV16 stem-loop ddPCR assay detects TR-ctDNA fragments in urine from patients with HPV⁺ OPSCC and shows concordance with results from plasma ctDNA analysis.

(A) Schematic of the short-amplicon (42 bp) stem-loop 2-stage PCR approach (17) used to detect and quantify ultrashort HPV16 TR-ctDNA present in urine (see Supplemental Figure 3 for details). (B) Analytical validation of the stem-loop assay using a 2-fold dilution series to define the detectable range (top plots). Expected copies of synthetic spiked-in HPV16 target DNA (left) or the GEs of HPV16⁺ cancer cell line UM-SCC-104 tested (right) were plotted against the measured copies (cumulative of triplicates). LoD was determined to be 4.2 copies (see Methods). Bottom plots show observed coefficient of variation (%CV) corresponding to measurement at each dilution of synthetic HPV16 DNA or UM-SCC-104 gDNA. (C) Stem-loop assay was used to detect HPV16 ctDNA in matched urine and plasma samples collected before treatment, from 32 HPV⁺ OPSCC patients representing both locally advanced (LA) and metastatic (M) cases, and 12 negative controls (1 HPV18⁺ OPSCC patient, 6 HPV^– head and neck squamous cell carcinoma [HNSCC] patients, and 5 healthy individuals without cancer). The assay detected HPV16 ctDNA in both urine and plasma in 27 of the 32 cases of HPV⁺ OPSCC. HPV16 ctDNA was not detected in any of the negative controls (12 of 12 cases), showing high concordance between plasma-based and urine-based ctDNA detection. (D) Comparison of log10 HPV16 ctDNA values in urine and matched plasma from 31 patients with HPV⁺ OPSCC (see Methods). Significant correlation was found between HPV16 copies detected in urine and plasma based on Pearson’s correlation test; *P (2-tailed) = 0.0014. HPV16 TR-ctDNA values (cumulative of ddPCR triplicates) correspond to the mean value of 30 mL urine samples from 2 bottles; and ~0.9 ml for plasma samples..

Figure 3. Proof of concept for earlier detection of cancer recurrence via serial urine HPV16 TR-ctDNA measurements in 4 patients with HPV⁺ OPSCC.

Shown are ddPCR results from testing of longitudinal urine collections from 4 patients, using the stem-loop 42 bp urine TR-ctDNA assay. Number of HPV16 copies detected (y axis) was plotted on a log₁₀ scale (patient 1 and patient 3) or linear scale (patient 2 and patient 4) as HPV16 ctDNA values (cumulative of ddPCR triplicates) from urine cfDNA samples collected at different time points over several months (x axis). Samples were analyzed from 7 time points for patient 1, 4 time points for patient 2, 4 time points for patient 3, and 3 time points for patient 4. Closed symbols represent HPV16 molecules detected above LoD (magenta dotted line), and open symbols represent values that were below the LoD, with serially collected samples at baseline prior to treatment in black, samples collected during treatment in green, and samples collected during posttreatment surveillance for recurrence in red. The day of treatment is marked as 0 on the x axis. Patients 1–3 underwent chemoradiation (CRT) treatment (7 weeks) and showed detectable HPV16 TR-ctDNA prior to initiation of treatment. Patient 4 underwent surgical resection but, notably, had baseline urine DNA of poor quality and, therefore, that time point could not be accurately analyzed for HPV16 TR-ctDNA. In 3 patients, HPV16 TR-ctDNA was detected in urine during the surveillance period prior to clinically detected recurrence.