Evaluation of urinary cfDNA workflows for the molecular profiling of malignant disease

Summary

Urinary cell-free DNA (ucfDNA) is a promising liquid biopsy analyte obtained by a non-invasive sampling method, particularly valuable in urological cancers due to the proximity of biofluids to the tumor site. However, extremely high fragmentation and low abundance of ucfDNA pose significant analytical challenges. Improper handling and storage can severely affect ucfDNA quality, necessitating optimized pre-analytical workflows for clinical applications. We evaluated multiple pre-analytical workflows, including urine stabilization approaches against unstabilized urine samples. Our results demonstrated that without stabilization, ucfDNA degraded rapidly and minimal quantities remaining after 72 h. In contrast, urine stabilization preserved ucfDNA integrity, enabling successful downstream analyses, including digital PCR and next-generation sequencing, for mutation detection in cancer-related genes and genome wide copy number profiling. These findings underscore the critical role of stabilization for a reliable ucfDNA analysis and highlight its potential in molecular diagnostics as a complementary tool to plasma-based liquid biopsies.

Graphical abstract

Highlights

• •Native urine rapidly degrades ucfDNA, impairing downstream molecular analyses
• •Urine stabilization preserves ucfDNA quantity and integrity for several days
• •Stabilized ucfDNA enables reliable downstream analysis, including digital PCR and NGS
• •With proper sample handling, ucfDNA can be a valuable source of genetic information

Diagnostic procedure; Fluid; Diagnostic technique in health technology

Introduction

Recent advances in liquid biopsy (LB) research have boosted the molecular profiling of tumor DNA and management of patients with cancer.¹ While initial LB research primarily focused on blood-derived tumor biomarkers, such as circulating tumor DNA (ctDNA), recent studies have identified cancer-derived biomarkers in other body fluids, including cerebrospinal fluid, pleural fluid, saliva, peritoneal fluid, and urine.² Of these, urinary cell-free DNA (ucfDNA) holds great promise, since urine collection is simple, truly non-invasive, and painless compared to tissue biopsies or blood draws, making it more acceptable and easier for patients to undergo repeated sampling. Unlike plasma, where cell-free DNA (cfDNA) primarily originates from hematopoietic cells,³ ucfDNA is thought to be derived from two distinct sources, i.e., cells of the urinary tract directly releasing DNA into urine and/or trans-renal DNA, which originates from plasma and enters the urine by passing the glomerular filtration system.⁴ Trans-renal DNA is typically of smaller size (<250 bp), because of the glomerular filtration barrier, which only allows smaller fragments to pass through. CfDNA derived from cells in the urogenital tract can be of various sizes, also including high molecular weight DNA.^{5,6,7,8,9,10,11}

Although proof-of-principle studies have indicated the utility of ucfDNA in non-urological cancers such as colorectal, non-small cell lung cancer and hepatocellular carcinoma,^{6,7,12,13,14,15} ucfDNA is particularly effective for urogenital cancers, such as bladder, prostate, and renal cancers, as it may contain tumor-specific alterations that are shed directly into the urine.^{16,17,18,19,20,21,22} In this context, several studies demonstrated that samples taken from locations close to the tumor site can enhance diagnostic sensitivity.^{9,23,24,25,26,27}

The feasibility of ucfDNA was shown for several applications, such as tumor-informed mutation analysis,²⁸ assessment of microsatellite instability,²⁹ or genome-wide copy number profiling.³⁰ However, ucfDNA analysis presents several challenges that must be addressed to optimize its utility in clinical diagnostics. A significant challenge is the harsh environment within urine, characterized by the variable and acidic pH (ranging from 5.0 to 7.0) and a high activity of nucleic acid hydrolyzing enzymes, which are 100-fold more active in urine compared to plasma.^4,31 The presence of nucleases and other enzymes in urine rapidly degrades cfDNA, complicating the extraction process and potentially compromising the integrity of cfDNA samples.^4,32,33 This results in lower overall concentrations of cfDNA in urine compared to blood, and a need for highly sensitive detection techniques to accurately identify and quantify tumor-derived DNA. Contamination is another critical issue, as urine can contain DNA from non-cancerous cells or bacteria.^34,35,36 All these factors can increase noise, interfere with the specificity of the detection of cancer-specific signals, and hence, lead to false-positive results. Moreover, inflammations may lead to the exfoliation of inflammatory cells, which leads to an increased amount of “normal” DNA, which may dilute tumor-derived signals and increase the false-negative rate.³³ Minimizing these effects is key for preserving the cfDNA profile and ensuring reliable ucfDNA analysis.

Methods for urine collection, processing, and ucfDNA analysis can significantly affect ucfDNA analysis. While the use of preservatives in blood collection tubes is generally accepted and recommended for cfDNA analyses from plasma, recommendations for ucfDNA are less advanced.³⁷ Moreover, the analytical workflow is an additional aspect that influences ucfDNA analysis. Therefore, variability in protocols across different laboratories can result in inconsistent findings, making it challenging to compare results and establish reliable diagnostic criteria.^38,39,40

Here, we evaluated pre-analytical workflows with and without stabilization in combination with several downstream analyses, including cfDNA quantification, mutation analysis, and genome-wide copy number profiling. More specifically, we tested two commercially available urine stabilization devices including the Streck Urine preservative (Streck) and a new urine collection and stabilization solution called PAXgene Urine Liquid Biopsy Set (PreAnalytiX) regarding the preservation of the original proportion and integrity of ucfDNA in self-reporting healthy controls and compared it to unstabilized urine (Figure 1A). In addition, we tested the compatibility of stabilized and unstabilized ucfDNA, respectively, with different downstream analyses, including digital PCR (dPCR) and next generation sequencing (NGS) approaches (Figure 1B).

Figure 1.

Study overview

(A) Urine samples were collected from self-reporting healthy controls, aliquoted and stored with (PAXgene, PAXgene Urine preservative or Streck, Streck Urine Preserve) and without stabilization (native). Urine was processed immediately (h0), and after 6, 24, 48, and 96 h and urinary cell-free DNA (ucfDNA) was extracted on a QIAsymphony. ucfDNA quantification was performed using the fluorometric system Qubit, a digital PCR (dPCR) RNAseP assay, a quantitative PCR assay amplifying the 18S RNA gene (RNA18S1), as well as the Quantiplex PCR employing TaqMan probes including a 91bp as well as a 353bp PCR product in an autosomal region (4NS1C) for the quantification of human DNA, an 81bp target region for the quantification male DNA, and a 434bp internal control. Samples below the limit of detection are plotted at 0.01.

(B) Matching plasma and urine samples were collected from breast, colorectal, and prostate cancer patients, aliquoted and stored with (PAXgene Urine preservative or Streck, Streck Urine Preserve) and without stabilization (native). Urine was processed on the same day (d0), and after three days (d3) and ucfDNA was extracted on a QIAsymphony. For each cancer type different down-stream analyses were applied including digital PCR for PIK3CA hotspot testing (dPCR), an amplicon sequencing approach, a hybrid capture-based enrichment, and shallow whole genome sequencing (sWGS). Created with Biorender.

Results

Urinary cell-free DNA in healthy controls

As a first step, we evaluated the effect of urine stabilization in self-reporting healthy volunteers (females n = 19; males, n = 21) at various storage durations (Figure 1A). When we analyzed the size distribution of native (unstabilized) ucfDNA using a TapeStation, we observed a main peak between 100 and 500bp in line with previous reports, which demonstrated that ucfDNA is more degraded, shorter, and variably sized in contrast to plasma cfDNA, demonstrating a non-random nucleosomal shaped size distribution.⁴¹ Already after 3 h of storage, an increase in the total amount of DNA was observed, which could be attributed to the release of gDNA from cells. At the same time, DNA fragments with smaller sizes increased, indicating the rapid degradation of DNA by nucleases, potentially starting immediately after donation (Figure 2A). After 6 h, a substantial increase of smaller fragments (<150bp) occurred, and after 24 h almost all larger fragments had disappeared. After 72 h, very few fragments remained, and cfDNA concentration decreased significantly.

Figure 2.

ucfDNA concentrations with and without stabilization

(A) Shown are TapeStation profiles of native (unstabilized) ucfDNA from an exemplary case extracted after 1, 3, 6, 24, and 72 h of storage before extraction demonstrating a rapid degradation of ucfDNA. Marker peak at 35bp.

(B) Plotted are ucfDNA concentrations assessed with Qubit from self-reporting healthy controls (females, n = 20; males, n = 21) and cancer patients (BrCa-breast cancer, n = 12; CRC-colorectal cancer, n = 15; PCa-prostate cancer, n = 20) from urine stabilized with the PAXgene Urine preservative. Two-tailed Mann Whitney test rank was used for comparisons.

(C) Shown are ucfDNA concentrations (ng/mL urine) from females and males extracted from urines stabilized with either the PAXgene Urine preservative prototype or without a preservative (native). Samples were processed immediately (h0), after 6 h (h6), 24 h (h24), 48 h (h48) and 96 h (h96). Quantification was performed using an RNAse P dPCR assay.

(D) Same as in (C) but urine was stabilized with either PAXgene or the Streck urine preservative. ns, not significant.

Since urine contains only minute amounts of cfDNA molecules, we intended to use several quantification methods to limit the technical variation and ensure reliable results. Generally, ucfDNA levels were significantly higher in females compared to males, irrespective of stabilization, storage duration, or quantification method (for Qubit at d0 using PAXgene 8.1 ng/ml urine vs. 1.7 ng/ml urine, with several samples below the detection limit in males, p < 0.001, two-tailed Mann Whitney test) (Figure 2B). However, ucfDNA concentrations in females showed slightly more intra-individual and time point differences (Figure S1), which might have physiological/anatomical reasons.^42,43 Not surprisingly, Qubit measurements revealed a higher variability within groups than the qPCR-based methods, which are more accurate at low concentrations (Figures S2 and S3). For both PAXgene and Streck stabilized urine, concentrations remained stable over time, whereas ucfDNA concentrations in native urine were significantly lower across all time points (p < 0.001, Wilcoxon matched-pairs signed rank test for dPCR-based quantification) (Figures 2C and 2D, S4). However, due to the overall low levels (in many cases close to or below the LOD) the differences in ucfDNA amounts between PAXgene and unstabilized samples were less pronounced in males (Figure 2C and 2D).

PIK3CA hotspot testing using digital PCR

In order to test whether a hotspot testing approach for ucfDNA was feasible, we recruited patients with HR + breast cancer (n = 14) who had previously undergone PIK3CA mutation testing, originally performed on plasma cfDNA using our in-house developed SiMSen-Seq assay.⁴⁴ First, we analyzed corresponding plasma and urinary cfDNA collected at the same time point using the same method. However, since the SiMSen-Seq assay was specifically optimized for plasma-derived cfDNA, library preparation failed for most ucfDNA samples - likely due to differences in fragment size distribution and lower amounts of amplifiable cfDNA in urine (Figure S5A).

Therefore, we employed a validated dPCR approach for plasma cfDNA and analyzed nine patients with either E545K or H1047R mutations. To this end, plasma cfDNA and all available ucfDNA samples with different stabilization and storage conditions collected at the same time point were analyzed using a duplex mutation assay (Table S1). In 5/9 plasma cfDNA samples (55.5%), the mutation previously identified with SiMSen-Seq could be reliably confirmed (Figure 3), whereas none of the urine samples yielded clearly positive results.

Figure 3.

Digital PCR for hotspot testing of PIK3CA mutations in cfDNA from plasma and urine in breast cancer

Shown is the number of mutated (Mut) and wild type (WT) copies from four exemplary cases assessed with a (A) PIK3CA H1047R and (B) PIK3CA E545K digital PCR assay. ucfDNA was extracted at the day of donation (d0) and after three days of storage (d3) from urine with (PAXgene, PAXgene Urine preservative prototype; Streck, Streck urine preservative) and without stabilization (native). The dotted red line indicates the limit of detection (LOD) of the assay.

While the number of available copies (quantified by the PIK3CA wild type assay) remained relatively constant from d0 to d3 for samples with either PAXgene or Streck stabilization, at d3 significantly fewer copies were available for native ucfDNA, confirming a rapid degradation (Kolmogorov-Smirnov test, PAXgene vs. Streck p = 0.9886, PAXgene vs. native p = 0.0097, Streck vs. native p = 0.0159) (Figure 3, and S5B). When using native urine at d0 as a reference, stabilized samples from four patients - each with all storage conditions available - showed 6- to 7-fold and 15- to 17-fold higher PIK3CA WT copy numbers at d0 and d3, respectively. Native urine, however, retained just 16% of the initial cfDNA at d3, indicating more than 80% degradation. Furthermore, when comparing copy numbers at d3 across conditions, stabilized samples exhibited an 81- to 87-fold higher PIK3CA WT copy number relative to unstabilized samples. When evaluating mutant fragments – all of which were close or below the LOD- samples from patients BC17 showed the consistent stabilization of mutant alleles, while BC15 and BC16 exhibited slightly reduced mutant signals under stabilized conditions.

Mutation detection homologous recombination repair genes using amplicon sequencing

To test whether a more comprehensive mutation analysis was feasible in ucfDNA, we used a custom QIAseq panel that enriches for 15 genes involved in homologous recombination repair (HRR) as well as RB1, PTEN, and TP53. Homologous recombination deficiency (HRD) testing in cfDNA from plasma has emerged as a valuable tool to identify patients who are likely to respond to PARPi.⁴⁵ First, we analyzed plasma and urine samples from patients with metastatic prostate cancer (n = 20), of which paired plasma cfDNA and Streck ucfDNA at d0 were available from a previous study collection (article in preparation) (Figure S6). At least one mutation was identified in 18/20 plasma cfDNA samples (90%). The same mutation could also be detected in the corresponding ucfDNA in 9/18 (50%) patients, but with substantially lower VAFs (Table S2). In a prospective collection of 10 additional patients, urine was collected and aliquoted as outlined in the method section. However, in most cases, the obtained volumes (on average 51.8 mL, range 24–110 mL) were insufficient to provide for all stabilization conditions of the study set-up. While library preparation was successful for all of the stabilized urine samples, for 6/8 (75.0%) of native ucfDNA samples, either library preparation failed or ucfDNA yield was too low to achieve the required input. Quality metrics were in the expected range for all samples with successful library preparation regardless of the analyte or stabilization reagent (Figure S7). In 4/10 (40%) patients, no mutation was detected in either plasma or urine. In patients with ctDNA detected, tumor fractions seemed higher in plasma cfDNA, but tumor-derived DNA could also be detected in ucfDNA samples (Table S2). In one patient (#P364) with sufficient urine volume for all storage conditions, a putative germline variant in BARD1 was identified in all samples with VAF around 50% whereas a somatic PTEN variant could only be identified in stabilized samples, but not in the native one (Figure 4A). For another patient (#P631), for whom only ucfDNA from PAXgene d0 and plasma were available, all three variants detected in plasma could also be identified in ucfDNA (Figure 4B). In patient #P637, a low-level mutation in BRCA2 could be identified in plasma and ucfDNA stabilized with PAXgene or Streck for three days, respectively. Moreover, in PAXgene ucfDNA, an additional mutation in RB1 was identified (Figure 4C). Taken together, even if tumor levels in urine were low in our patients - potentially leading to sampling biases between the various aliquots -, stabilized urine yielded high quality sequencing data using an amplicon based-approach, while native urine could not be successfully analyzed using the QIAseq panel.

Figure 4.

Mutation detection in homologous recombination repair (HRR) genes from ucfDNA using a custom amplicon panel in prostate cancer

Shown are three exemplary cases of corresponding ucfDNA and plasma cfDNA samples from prostate cancer patients

(A) For this patient sufficient urine was available to test all storage conditions (d0, urine processing on the same day; d3, processing after three days; native, no stabilization; PAXgene Urine preservative prototype; Streck, Streck urine preservative). A BARD1 germline mutation was identified in all samples, while a somatic PTEN mutation could only be identified in stabilized ucfDNA.

(B) For this patient only PAXgene ucfDNA was available, but a BRCA2 germline mutation as well as two somatic mutations (PTEN and TP53) could be identified in both plasma cfDNA and ucfDNA.

(C) In this patient, the same BRCA2 mutation was identified in plasma cfDNA and ucfDNA stabilized with PAXgene or Streck. In PAXgene ucfDNA an additional mutation in RB1 was observed.

Mutation detection in clinically relevant genes using a hybrid capture based sequencing approach

To test a hybrid capture based sequencing approach, we prospectively recruited patients with metastatic CRC (n = 10) and collected blood and urine samples at the same time point (Table S3). As with the other recruitments, in many cases insufficient urine volume was donated to enable the testing of all stabilization and storage conditions in parallel. While hybrid capture based enrichment yielded high quality libraries for all plasma cfDNA and stabilized ucfDNA samples, respectively, library preparation failed in 60.0% (3/5) of native urine samples. In the four native ucfDNA samples, for which sequencing data were available, the unique sequencing depth was substantially reduced, and after three days of storage, the error rate was higher compared to plasma and stabilized urine samples (Figure S8). It is noteworthy that also stabilized ucfDNA samples had increased mean error rates compared to plasma cfDNA (Figures 5A and 5C). This was also reflected by a larger number of low-level variants (VAF<0.5%), particularly if less than 50 ng was available as input amount. Median fragment length analysis revealed the expected mean fragment length of approximately 166bp for the plasma sample from patient C412, but was slightly higher for patient C430. For urine samples, a broader range was observed (Figure S8).

Figure 5.

Mutation detection in ucfDNA from colorectal cancer patients using a hybrid capture based enrichment in colorectal cancer

Shown are error rates and sequencing results from two exemplary patients (C430, upper panel; C412, lower panel) from corresponding plasma cfDNA and ucfDNA samples (d0, urine processing on the same day; d3, processing after three days; PAXgene, PAXgene Urine preservative prototype; Streck, Streck urine preservative; d0). Plotted are the (A and C) estimated error rates calculated from the analysis pipeline, and (B and D) the detected variants in each sample and their variant allele frequencies (VAF). The dotted red line indicates the limit of detection (LOD) of the assay.

Putative germline variant calls from all samples of the same patient revealed full concordance. Except for four cases, at least one putative somatic variant was identified in plasma cfDNA, while detections rate in ucfDNA were lower when applying the detection threshold of 0.5% (Table S4). In some cases, the same variants identified in plasma cfDNA were detected in ucfDNA, but always with lower VAFs (Figure 5B). However, a variety of variants unique to only one sample were detected in the majority of ucfDNA samples, which might still represent technical errors. In one patient, several mutations in canonical CRC driver genes were detected in both PAXgene and Streck ucfDNA with VAF below the LOD but not in the corresponding plasma cfDNA, which might represent true variants and indicated that complementary information may be retrieved in urine from patients with CRC (Figure 5D).

Shallow whole genome sequencing (sWGS) for copy number profiling and fragment size analysis

To investigate the effect of urine stabilization on sWGS, ucfDNA from patients with PCa (n = 12) was analyzed (one with a follow-up time point), of which, except for two patients, sufficient urine was donated to allow testing of all storage conditions. In 3/12 plasma ucfDNA samples (25%), tumor-specific somatic copy number alterations (SCNA) could be detected, whereas all others had tumor fractions below the limit of detection. Of those, two also had SCNA detected in corresponding ucfDNA samples. In patient #P364, PAXgene and Streck urine showed highly similar tumor fractions, while in native urine (d3), the tumor fraction was slightly decreased (Figure 6). Interestingly, in a follow-up sample from the same patient, the tumor fraction increased in urine but decreased in plasma (Figure S9). In patient #P631, all urine samples had tumor fractions of around 20% irrespective of the storage condition or type of stabilization, which indicates that sWGS is a robust method that is less impaired by the high fragmentation of ucfDNA (Figure S10).

Figure 6.

Genome-wide copy number profiling of plasma and urinary cfDNA with and without stabilization in prostate cancer

Shown are copy number profiles and estimated tumor fractions of urinary and plasma cfDNA from a prostate cancer patient establisher from shallow whole genome sequencing data (d0, sample processing on the same day; d3, processing after three days; plasma, cfDNA from blood plasma; PAXgene Urine preservative prototype; Streck, Streck urine preservative, native, no stabilization).

To assess the effect of urine stabilization on ucfDNA fragment size distributions, we additionally sequenced ucfDNA from healthy individuals (n = 13). Fragment size distribution for ucfDNA calculated from paired-end sWGS data was in line with previously published reports showing fragments in the range of 50-500bp with a typical 10bp periodicity (Figure 7).

Figure 7.

Genome-wide ucfDNA fragment patterns depend on stabilization and processing time

Comparison of the mean urine cfDNA size distribution determined by shallow whole genome sequencing data from healthy individuals (left panel) and prostate cancer patients (right panel). Size profiles from matched PAXgene stabilized (yellow), Streck-stabilized (blue), and unstabilized samples (gray) are shown. The upper panel displays the fragmentation patterns at day 0 (d0), while the lower panel shows fragmentation at day 3 (d3).

While ucfDNA stabilized with PAXgene and Streck exhibited highly concordant fragment size profiles, native ucfDNA showed a higher degree of fragmentation already at d0, which became even more pronounced after 3 days of storage (d3). In native urine samples at d0, an average of 30.8% and 25.9% of fragments were below 150bp in controls and patients with cancer, respectively. These proportions increased to 36.5% and 28.2% at d3. In contrast, ucfDNA from PAXgene and Streck-stabilized samples showed comparable and lower levels of short fragments at both time points, i.e., 24.8% and 24.4% at d0, and 21.2% and 23.0% at d3, respectively. Interestingly, these differences were less evident in ucfDNA from patients with PCa cancer compared to healthy donors. Analysis of fragmentation abundance per 50bp bins highlighted a consistent shift toward shorter fragment sizes in unstabilized samples (Figure S11).

Discussion

Unlike plasma cfDNA, for which pre-analytical variables needed for a successful translation into clinical practice are well defined,^46,47,48 there is no broad consensus regarding the requirements for ucfDNA analysis. Although extraction methods specifically designed for small fragments have been shown to be efficient for ucfDNA isolation,⁴⁹ most of the existing studies on ucfDNA have not provided specifics for urine stabilization.⁵⁰ In some studies, EDTA has been used as a preservative,^41,51,52 but in recent years, commercial preservatives dedicated to urine stabilization have become available. Here, we used the Streck urine Preservative and tested the recently launched PAXgene Urine Liquid Biopsy Set for ucfDNA, which, in contrast to other systems, allows for straightforward urine collection and stabilization with a standardized ratio of urine to stabilizer embedded in a complete pre-analytical workflow from urine collection to cfDNA isolation. First, we tested urine cfDNA stabilization in self-reporting healthy volunteers and compared ucfDNA quantities and integrities to unstabilized urine. Although manufacturers-reported stabilization periods are up to 7 days for Streck and 10 days for PAXgene, our study focused on short-term urine storage (up to 4 days at room temperature). While different downstream analytical methods were applied across cohorts, pre-analytical procedures - including urine collection, stabilization, and storage - were standardized throughout the study. To control for biological variation, second morning urine was used in healthy volunteers, allowing for standardized sampling and immediate processing. In contrast, clinical samples were collected throughout the day, underscoring the importance of effective stabilization regardless of collection time.

Without stabilization, ucfDNA degrades rapidly, with significant fragmentation occurring as early as 3 h post-collection, leading to significant changes in fragment length distribution and a marked reduction in detectable DNA fragments over time. Consistent with previous reports, we found that ucfDNA levels were significantly higher and exhibited greater intra-individual and temporal variability in females compared to males, regardless of stabilization, storage duration, or quantification method.^43,53 This difference might be attributed to physiological or anatomical factors and the fact that female urine may contain more epithelial cells than that from males.

The consistently higher ucfDNA concentrations in stabilized urine samples, even at baseline (d0), likely reflect protection against the rapid degradation that occurs in native urine immediately after collection. This is supported by sWGS data showing greater fragmentation and reduced cfDNA complexity in native samples. Moreover, previous spike-in experiments have similarly shown lower recovery of exogenous cfDNA from native urine, even when processed immediately, supporting the notion of compromised preservation.⁵³ Nevertheless, immediate effects of stabilizers - such as partial cell lysis or chemical interactions - may also contribute to elevated DNA levels. However, the rapid degradation kinetics in native urine complicate its use as a reference. Future studies should investigate the impact of stabilizer-to-urine ratios and potential interactions with cfDNA assays (e.g., effects on DNA post-extraction). These factors are critical to consider when interpreting yield differences and designing standardized pre-analytical workflows.

The observed degradation pattern of ucfDNA, especially the shift toward smaller fragments in unstabilized urine, suggests that ucfDNA is strongly susceptible to external factors such as enzymatic degradation. Stabilization with PAXgene and Streck, respectively, maintained the integrity and quantity of ucfDNA, and although not significant, ucfDNA yields were slightly higher when PAXgene was used.

Fragment size analysis from sWGS data of healthy individuals and patients with prostate cancer supported previous reports, showing that ucfDNA fragments are generally within the 50-500bp range, with a characteristic 10bp periodicity. The stabilization of urine samples with both PAXgene and Streck preserved this size profile, whereas unstabilized samples showed increased fragmentation, especially in healthy donors. This suggests that urine stabilization not only preserves DNA integrity but also maintains the representativeness of the DNA fragments, which is crucial for accurate copy number profiling and other genomic analyses, particularly fragmentomics approaches.⁴¹

In dPCR analyses, mutated PIK3CA fragments were detected in some ucfDNA samples; however, these did not reach the threshold for a positive call. Notably, in unstabilized urine samples, the number of amplifiable cfDNA fragments declined significantly after three days of storage at room temperature, which would markedly reduce the sensitivity of mutation detection. The variability in PIK3CA mutation detection across samples (Figure 3) is likely attributable to stochastic effects at low VAFs near the assay’s limit of detection (LOD), rather than to shortcomings in the stabilizers or assay performance. Variability is to be expected when working near the detection limit, and these findings should be interpreted as a proof-of-principle for the feasibility of mutation detection in urine cfDNA, rather than as evidence of a fully optimized diagnostic workflow for locally advanced breast cancer.

Importantly, it should also be noted that, on average, 12.8 months (range: 0–29.5 months) had passed between the initial mutation detection and the subsequent blood sampling and urine donation used in this study, which may explain differences in detection rates. All included patients had locally advanced tumors, and some were receiving treatment at the time of sampling, likely contributing to reduced ctDNA shedding in plasma - the original source of mutation detection - and, consequently, to low tumor-derived cfDNA fractions in urine. In general, relatively little is known about the presence and dynamics of tumor-derived DNA in urine in breast cancer. However, one study in triple-negative breast cancer demonstrated that urine and plasma can offer complementary insights into the tumor’s genetic profile, suggesting potential value in parallel analysis of both biofluids.⁵⁴

Our data also demonstrate that the low abundance and high fragmentation of ucfDNA can pose technical challenges in library preparation. Although sWGS for copy number profiling was possible for all samples, our attempts to detect mutations in ucfDNA encountered challenges due to the fragmented nature and lower quantity of ucfDNA compared to plasma cfDNA. For the majority of native ucfDNA samples, insufficient input material was available or library preparation failed for both amplicon- and hybrid-capture-based enrichment. For the few native ucfDNA samples, in which sequencing data could be generated, data quality was inferior compared to the stabilized samples. It is worth noting that error rates of stabilized ucfDNA were also slightly higher than those of plasma cfDNA. Most likely, inefficiencies in library conversion introduced biases and led to a higher error rate. Moreover, bioinformatics pipelines used to analyze NGS data may struggle more with the noisier and lower-quality data from ucfDNA, leading to higher error rates in variant calling and other analyses. The higher background noise can mask somatic mutations and other genetic variations, thereby reducing the accuracy of NGS approaches. In addition, mapping short, fragmented sequences to the reference genome is more prone to errors, which can be exacerbated by the lower quality of ucfDNA. In summary, the higher error rate of NGS of ucfDNA compared to plasma cfDNA is caused by lower quality, higher fragmentation, and greater contamination of DNA found in urine, as well as technical and bioinformatics challenges associated with sequencing and analyzing this type of sample.

Nevertheless, in properly stabilized cases where sufficient ucfDNA was available, we successfully detected mutations in several cancer-related genes using targeted sequencing panels. This indicates that ucfDNA could serve as a non-invasive biomarker source for specific mutations with proper stabilization and handling. Both PAXgene and Streck reliably stabilized urine and enabled all tested downstream analyses. It is of note though, that due to our study design, which required multiple aliquots from one urine donation, in many cases, less than 10 mL of urine was used for ucfDNA extraction. In a clinical framework, substantially higher volumes could be used, leading to higher yields and, in turn, higher quality of sequencing data. In addition to pre-analytical conditions, library preparation techniques and bioinformatics pipelines need to be improved to mitigate the aforementioned issues and enhance the accuracy of ucfDNA sequencing in the future.

From a clinician’s perspective, the ease of handling makes the PAXgene device particularly well suited for sample collection in clinical routine. A closed stabilization system, such as the PAXgene Urine Liquid Biopsy Set, offers significant practical advantages, as open handling of chemicals is not ideal in clinical settings where fast turnaround times and biosafety are essential. Unlike the Streck preservative, which requires manual mixing and volume adjustments, the PAXgene system uses a vacuum-filled design that automatically ensures a consistent urine-to-preservative ratio, eliminating variability and reducing labor. In our study, we manually maintained a fixed urine-to-Streck ratio to ensure consistency across samples; however, this approach is not feasible in routine clinical workflows. Accordingly, the standardized format of the PAXgene device was preferred by clinical staff. Seamless integration into clinical practice is further supported by the familiarity of the PAXgene system among healthcare professionals and patients, as successfully demonstrated in a real-world clinical setting during the H2020 project Instand-NGS4P (article in preparation). Given the importance of immediate nuclease inactivation and bacterial growth suppression to preserve native ucfDNA profiles, user-friendly, bedside-compatible collection devices are essential for robust and reproducible urine-based workflows. Moreover, closed, user-friendly systems such as the PAXgene device support safe, reliable, and error-minimized sampling outside clinical settings, making them well suited for decentralized testing and longitudinal monitoring, offering a great potential for at-home collection.

Overall, our study underscores the critical importance of urine stabilization in ucfDNA research and its potential applications in clinical diagnostics. While challenges remain, particularly in mutation detection due to the fragmented and low-abundance nature of ucfDNA, our findings suggest that with proper sample handling, ucfDNA can be a valuable source of genetic information. Particularly in urogenital cancers, the sensitivity of liquid biopsy could be improved by ucfDNA analyses, provided that standardized pre-analytical workflows are used. Future studies should focus on optimizing methodologies for ucfDNA analysis to fully realize its potential in non-invasive diagnostics and monitoring of various diseases. Moreover, it would be valuable to evaluate the reliability of urine-based cfDNA analysis across different tumor types and anatomical contexts, as this could inform more tailored liquid biopsy strategies.

Limitations of the study

One limitation of our study is that the detection rates in BrC and CRC were low, which indicates that, due to its higher tumor fraction, plasma cfDNA may be more reliable for mutation detection in non-urogenital cancers compared to ucfDNA. However, the scope of this study was a technical evaluation of downstream applications for ucfDNA rather than a direct comparison of plasma cfDNA and ucfDNA.

Resource availability

Lead contact

Requests for additional information and resources should be directed to the corresponding author, Ellen Heitzer (ellen.heitzer@medunigraz.at).

Materials availability

This study did not generate unique reagents.

Data and code availability

The data that support the findings of these studies are available from the corresponding authors upon request. Sequencing raw data have been deposited at the European Genome-phenome Archive (EGA; http://www.ebi.ac.uk/ega/), which is hosted by the EBI, under the accession number EGAS50000001093, including three datasets and are publicly available as of the date of publication (HRR gene mutation analysis; sWGS of cfDNA from plasma and urine; AVENIO mutation analysis). Accession numbers are listed in the key resources table.

Acknowledgments

The analysis was supported by the Austrian Federal Ministry for Digital and Economic Affairs (Christian Doppler Research Fund for Liquid Biopsies for Early Detection of Cancer with PreAnalytiX as a commercial partner). The CD model enables cooperation between science and business that is meaningful, useful, and productive both for the participating partners and for society. The work of PMA was supported by the H2020 PCP project Instand-NGS4P (Grant Agreement #874719).

Author contributions

EH, DM, TV, TK, MS, and PMA conceived and supervised the study; TB, MB, ND, and AG provided plasma and urine samples. AE, LZ, LG, GV, and RG performed experiments. EH, AE, TM, and MM performed data analysis and visualization. AB supervised the work on the project. ML contributed to project administration. EH acquired funding. AE, TM, and EH drafted the article. TV, DM reviewed and edited the article. All authors have read and approved the final article.

Declaration of interests

E.H. has received unrelated funding from Illumina, Roche, Servier, Freenome, and PreAnalytiX, and received honoraria from Roche, Incyte, Astra Zeneca for advisory boards unrelated to our study. M.S., T.K., T.V., and DM are employees of PreAnalytiX/QIAGEN GmbH, Hilden, Germany. Thorsten Voss participates in QIAGEN’s regular long term incentive program (LTIs).

Declaration of generative AI and AI-assisted technologies in the writing process

During the preparation of this work, the authors used ChatGPT for language editing. After using this tool, the authors carefully reviewed and revised the content as needed and take full responsibility for the content of the publication.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Biological samples
Urine samples from healthy individuals and cancer patients including breast cancer, CRC, and prostate cancer	Medical University of Graz, Austria	https://www.medunigraz.at/
Blood samples from cancer patients including breast cancer, CRC, and prostate cancer	Medical University of Graz, Austria	https://www.medunigraz.at/
Critical commercial assays
PAXgene Urine Liquid Biopsy Tube prototype	QIAGEN, Hilden, Germany	Cat. No. / ID: 769143
Streck urine preservative	HiSS Diagnostics	Product No.: 230631
QIAsymphony DSP Circulating DNA Kit	PreAnalytiX, Hombrechtikon, Switzerland	Cat. No./ID: 937556
AVENIO ctDNA Targeted Kit	Roche Diagnostics, Rotkreuz, Switzerland	Material Number: 09733736001
TruSeq DNA Nano Sample Preparation Kit	(Illumina, San Diego, CA, USA)	FC-121_4003
Deposited data
HRR gene mutation analysis	This paper	EGAD50000001586
sWGS of cfDNA from plasma and urine	This paper	EGAD50000001587
AVENIO mutation analysis	This paper	EGAD50000001588
Software and algorithms
QuantStudio 3D AnalysisSuite Local Appliance Software	Thermo Fisher Scientific, Waltham, USA	https://www.thermofisher.com/at/en/home/technical-resources/software-downloads/quantstudio-3d-analysissuite-local-appliance-software.html
QIAcuity Software Suite (version 2.2.0.26)	QIAGEN, Hilden, Germany	https://www.qiagen.com/us/resources/resourcedetail?id=a88b8c8e-0d0b-4e0b-8b8f-f22bb1e1935f&lang=en
AVENIO Oncology Analysis Software v2.0.0.	Roche Diagnostics, Rotkreuz, Switzerland	https://sequencing.roche.com/us/en/products/product-category/ngs-oncology-assays.html
Cutadapt (version 4.2)	M. Martin 2011. Cutadapt removes adapter sequences from high-throughput sequencing reads (60)	https://cutadapt.readthedocs.io/en/stable/
Python - seaborn (version 0.13.2)	© 2012-2024, Michael Waskom.	https://seaborn.pydata.org/
matplotlib (version 3.8.4).	J. D. Hunter. Matplotlib: A 2D Graphics Environment, Computing in Science55	https://pypi.org/project/matplotlib/3.8.4/
GraphPad Prism (version 9.4.1.)	GraphPad	https://www.graphpad.com/?ref=w3use
Oligonucleotide
PIK3CA E545K GeneGlobe	QIAGEN, Hilden, Germany	No. DMH0000292-B200
PIK3CA H1047R GeneGlobe	QIAGEN, Hilden, Germany	Cat. No. DMH0000036-A200

Experimental model and study participant details

Human subjects - donor recruitment

All donors were recruited within the framework of approved studies from the Ethics Committee of the Medical University of Graz. Urine samples were collected from self-reporting healthy volunteers (females, n=19; males, n=21, media age 30 years; range 27-47 years) (ethical vote 31–065 ex 18/19). Furthermore, matched blood and urine samples were collected from patients diagnosed with ER-positive breast cancer (BrC) with known PIK3CA mutations from prior cfDNA testing using the SiMSen-Seq method⁴⁴ (patients n=13, samples n=14, all female; median age 64years; range; 38-83 years) (ethical vote 32-415 ex 19/20), colorectal cancer (CRC) (patients n=10; female, n=9, male n=1; mean age 66 year; range 49-73 years) (ethical vote 21-229 ex 09/10), and metastatic prostate cancer (PCa) (patients n=30 for HRR; mean age 70years; range 54-88 years), n=12 for sWGS; mean age 71 years; range 51-82 years) (ethical vote 21228 ex 09/10). Written informed consent was obtained from all patients and healthy individuals, respectively. All participants self-identified as White (Caucasian). The study was conducted according to the Declaration of Helsinki and all analyses were performed in accordance with the guidelines for good scientific practice as required by the Medical University of Graz.

Method details

Urine and blood collection

From self-reporting healthy volunteers, 20-120ml of second-morning urine was collected into BD Vacutainer Urine Collection Cups (Becton Dickinson and Company, Franklin Lakes, USA) and then quickly split into 10ml aliquots. The first study set-up included a comparison of the newly developed PAXgene Urine Liquid Biopsy Tube prototype (PreAnalytiX, Hombrechtikon, Switzerland) containing a proprietary stabilization reagent (hereinafter referred to as PAXgene), and unstabilized urine (hereinafter referred to as native) after immediate processing (h0) and after storage at room temperature for 6 (h6), 24 (h24), 48 (h48), and 96 (h96) hours, respectively. In the second study set-up, the PAXgene prototype was compared to the Streck urine preservative (hereinafter referred to as Streck) (Streck, La Vista, NE, USA). While the PAXgene preservative is prefilled in vacutainer tubes (10ml) and provides an evacuated closed system that ensures a consistent urine-to-preservative ratio if used in combination with the PAXgene Urine Collection Cup, the Streck preservative had to be added to a urine sample manually in an open urine handling step and the ratio of urine and stabilizer varies depending on the collected urine volume. For this study, we decided on a ratio of 1:5 stabilizer to urine.

Blood and urine samples from cancer patients were collected during their routine visits to the oncological outpatient clinic at the Division of Oncology. Blood was collected in PAXgene® Blood ccfDNA Tubes (PreAnalytiX, Hombrechtikon, Switzerland). Urine was donated into PAXgene Urine Collection Cup prototypes. If sufficient volume was available, urine was aliquoted in 10ml portions, of which two remained unstabilized, two were stabilized with the Streck preservative and two were stabilized with PAXgene. Of each pair, one aliquot was processed immediately after delivery to the lab (d0, ∼0.5-2 hours), whereas the other aliquot was stored at RT for three days (d3). When urine volume was limited, we prioritized PAXgene over Streck over native, and d3 over d0.

Sample processing and cfDNA isolation

Plasma was isolated from whole blood by two consecutive centrifugations at 1900 × g for 10 min, with the supernatant being transferred to a new tube after each centrifugation. The resulting plasma was then aliquoted in 2 mL tubes and stored at -80°C until further processing. cfDNA was isolated from 2-4 ml using either the QIAamp Cirulating Nucleic Acid Kit (QIAGEN) according to the manufacturer's instructions or QIAsymphony PAXgene® Blood ccfDNA Kit (QIAGEN, Hilden, Germany).⁴⁴ The cfDNA concentration was measured using Qubit dsDNA High Sensitivity Kit (Thermo Fisher Scientific) and stored at -20°C until further analysis.

Urine samples (10 ml aliquots) from healthy volunteers were centrifuged at 1600 x g for 10 minutes (min) at room temperature (RT) and supernatants were stored at -80°C until further processing. ucfDNA was isolated from 4ml urine supernatant using the QIAsymphony DSP Circulating DNA Kit on a QIAsymphony SP instrument (QIAGEN, Hilden, Germany). After a small pilot study, the centrifugation protocol was adjusted for cancer patients. PAXgene and native urine samples were centrifuged at 1900 x g for 15min. The supernatant was transferred into a new 15ml tube and centrifuged for another 10min at 1900 x g. The Streck urine samples were centrifuged at 2700 x g for 10min according to Van Casteren et al. 2020.⁵⁶ ucfDNA was isolated either from 9.6 mL urine using QIAsymphony PAXgene® Blood ccfDNA Kit (PreAnalytiX, Hombrechtikon, Switzerland) or from 10ml urine using QIAsymphony DSP Circulating DNA Kit (QIAGEN, Hilden, Germany). ucfDNA from CRC samples was isolated from 2 x 10ml urine, then pooled and concentrated using the MinElute 96 UF PCR Purification Kit (QIAGEN, Hilden, Germany) following the manufacturer's instructions, to meet the high cfDNA input requirements for downstream analysis.

Quantification of ucfDNA using Qubit

ucfDNA was quantified using the dsDNA High Sensitivity Kit according to the user guide (Thermo Fisher Scientific, Waltham, USA), a fluorometric method. The Qubit dsDNA HS Assay Kit is highly selective for double-stranded DNA and was used according to the user guide “Qubit dsDNA HS Assay Kits” from Thermo Fisher Scientific. Briefly, the Qubit dsDNA HS Reagent was diluted 1:200 in Qubit dsDNA HS Buffer. This working solution was used to dilute the urine samples (1:100, 2 μl ucfDNA + 198 μl working solution) as well as two standards (1:20, 10 μl standard + 190 μl of working solution.) to a final volume of 200 μl. Each sample was mixed by vortexing 2 – 3 seconds and subsequently incubated at room temperature for 2 minutes before the measurement of the concentration.

Quantification of ucfDNA 18S rRNA PCR

A 66 bp fragment of 18S rRNA gene was quantified on the Rotor Gene Q 5plex HRM platform (QIAGEN, Hilden, Germany), whereas Human Female DNA was used for the standard series (Promega, Madison, USA). Briefly, 1 μl of PPM was mixed with 10 μl of QuantiTect Multiplex PCR master mix (QIAGEN), 6 μl of RNase free water and 3 μl of template (ucfDNA). The ucfDNA was diluted 1:2 with nuclease free water. The following program was run on a Rotor-Gene Q series 5plex HRM platform (QIAGEN) for the PCR reaction: 95°C for 15 min, followed by 40 cycles of 94°C for 60 sec and 60°C for 90 sec. Promega Human Female DNA was used for the standard series.

Quantification of ucfDNA using digital PCR

For accurate quantification of ucfDNA, a TaqMan RNase P assay in combination with the QuantStudioTM 3D Digital PCR Instrument was used (Thermo Fisher Scientific, Waltham, USA). To this end, 4 μl of the ucfDNA samples were mixed with 9 μl QuantStudio 3D Digital PCR Master Mix (Thermo Fisher Scientific) and 1 μl of the TaqMan RNase P assay. The mix was loaded into the applicator and distributed evenly by the chip loader. The PCR was run on a ProFlex Flat PCR System with the following program: stage 1: 96°C for 10 min, followed by 44 cycles of 56°C for 2 min and 94°C for 30 sec and stage 3: 10°C for 2 min. Image caption and preliminary analysis following thermal cycling was performed on the QuantStudio 3D Digital PCR Instrument. Final analysis of the imaging data was done via QuantStudio 3D AnalysisSuite Software. The copies/μl were converted to pg/μl with the following formular:

$$\frac{pg}{\mu l}=\frac{copies}{\mu l}x\frac{3200000000bpx600\frac{g}{mol}}{6.0221x{10}^{23}particles}x{10}^{12}$$

Quantification of ucfDNA using the Quantiplex pro PCR assay

The Investigator Quantiplex Pro multiplex assay combines the quantification of human genomic DNA, male DNA and the integrity of DNA in one tube and consists of four different TaqMan probes (QIAGEN, Hilden, Germany). Two probes targeting small DNA fragments; a 91bp autosomal fragment for the quantification of human DNA, and a 81bp target region on the Y-chromosome for the quantification of male DNA. Another autosomal fragment of 353bp selectively quantifies larger DNA molecules and a 434bp non-human fragment serves as internal control. The ucfDNA samples from cancer patients were quantified with Qubit only (Thermo Fisher Scientific, Waltham, MA, USA). The Quantiplex PCR was performed and analyzed according to QIAGEN, “Investigator® Quantiplex® Pro Handbook.” Mar-2018. For the PCR reaction 9 μl of Quantiplex Pro Reaction Mix were mixed with 9 μl of Quantiplex Pro Primer Mix and 2 μl unknown sample. Thermal cycling conditions were as follows 95°C for 3 min followed by 40 cycles of 95°C for 5 sec and 60°C for 35 sec.

Digital PCR for PIK3CA hotspot testing

To screen for PIK3CA mutations in ucfDNA, QIAcuity Multitarget LNA Mutation Detection Assays (QIAGEN, Hilden, Germany) were used (PIK3CA E545K GeneGlobe Cat. No. DMH0000292-B200 and PIK3CA H1047R GeneGlobe Cat. No. DMH0000036-A200). dPCR reactions for the QIAcuity One 5-plex instrument (QIAGEN, Hilden, Germany) were prepared according to the manufacturer's instructions (“HB-2819-001_LL_QIAcuityMutDet_0720_WW_v1Protokoll: Quick-Start Protocol July 2020 dPCR LNA® Mutation Assay”). Briefly, 10 μl of 4x QIAcuity Probe Master Mix were mixed with 1.33 μl 30x Primer/Probe Mix (E545K Atto550/Rox), 1.33 μl 30x Primer/Probe Mix (H1047R FAM/HEX), 0.10 μl Restriction Enzyme Xbal (0.025 U/μl), 2.33 μl dH2O and 25 μl of template to reach a final volume of 40 μl per reaction. Human gDNA mixed with gBlocks harboring the E545K and the H1047R mutation (IDT, Newark, USA) were used as positive controls and dH2O was used as negative control. The samples were incubated at room temperature for 10 min (enzymatic digestion) before running the PCR on the QIACuity One 5-plex instrument (QIAGEN, Hilden, Germany) with the following cycling conditions: initial PCR heat activation for 2 min at 95°C, followed by 40 cycles of denaturation for 15 s at 95°C and combined annealing/extension for 30 s at 60°C. Data analysis was performed using the QIAcuity Software Suite version 2.2.0.26.

Amplicon sequencing of HRR genes

Mutations in homologous recombination repair (HRR) genes were assessed in plasma and urine samples from 30 metastatic PCa patients using a customized QIAseq panel (QIAGEN, Hilden, Germany). The panel was designed to specifically enrich genes associated with homologous recombination repair deficiency (HRD) including ATM, BARD1, BRCA1, BRCA2, BRIP1, CDK12, CHEK1, CHEK2, FANCL, PALB2, PPP2R2A, RAD51B, RAD51C, RAD51D, RAD54L and customized to additionally enrich for RB1, PTEN and TP53. Library preparation was performed according to manufacturer’s recommendations, starting from 10-40ng of (u)cfDNA and quality controlled on a 2100 Bioanalyzer system (Agilent Technologies, Santa Clara, USA). Libraries were pooled equimolarly and then quantified by qPCR. Sequencing was performed on a NextSeq 550 or NovaSeq 6000 system, in 2x150bp paired-end mode, according to manufacturer’s instructions (Illumina, San Diego, USA). FASTQ files were analyzed using the GeneGlobe analysis platform (QIAGEN, Hilden, Germany), which includes adaptor trimming, short read removal and alignment using the Burrows-Wheeler algorithm (BWA-MEM version 0.7.9a-r786) on the reference genome hg19/GRCh37. Unique molecular identifiers (UMIs) were collapsed and variant calling was performed using the smCounter2 algorithm.⁵⁷ Variant annotation was performed with the Golden Helix VarSeq 2.2.0 (Golden Helix Inc., Bozeman, USA).

Hybrid capture based enrichment using the AVENIO ctDNA Targeted Kit

The AVENIO ctDNA Targeted Kit (Roche Diagnostics, Rotkreuz, Switzerland) enriches for 17 clinically relevant genes and is specially designed for lung cancer and CRC research. Libraries were prepared from an average of 39ng (u)cfDNA (range, 4.6-136.5), according to manufacturer's instructions. Library quantification was performed using the Qubit™ 1X dsDNA High Sensitivity Kit (Thermo Fisher Scientific, Waltham, USA) and quality was assessed on a Bioanalyzer High Sensitivity Kit (Agilent Technologies, Santa Clara, USA). The libraries were pooled by mass according to the Qubit dsDNA High Sensitivity Kit (Thermo Fisher Scientific, Waltham, USA) and quantified for sequencing via qPCR. Sequencing was performed in a paired end mode (2x150bp) on a NextSeq 500 or NovaSeq 6000 system (both Illumina, San Diego, USA). Analysis of the sequencing results was performed using the AVENIO Oncology Analysis Software v2.0.0. Concordance analysis of putative germline variants was performed to assess data quality. However, for the final analysis germline variants as well as variants supported by less than 10 alternative reads or below a VAF of 0.5% were filtered.

Copy number profiling using shallow whole genome sequencing

Shotgun libraries for shallow whole genome sequencing (sWGS) were prepared using the TruSeq DNA Nano Sample Preparation Kit (Illumina, San Diego, CA, USA).⁵⁸ If available, 10ng cfDNA and for ucfDNA 20ng input was used. If less was available, the maximum volume was used. Shotgun libraries were prepared following the manufacturer's instructions without the fragmentation step and 20-25 PCR cycles. Quality check was done on a Bioanalyzer DNA 7500 chip (Agilent Technologies, Santa Clara, USA). Libraries were pooled equally, quantified via qPCR and sequenced on the Illumina NextSeq or NovaSeq 6000 platform in paired-end mode (2x75bp or 2x151bp) (Illumina, San Diego, USA). Analysis of tumor-specific somatic copy number alterations (SCNAs) and tumor fraction estimation was performed using the ichorCNA algorithm.⁵⁹ Several adjustments were made to the default settings, including i) the use of an in-house panel of normal (n=17 for urine samples, and n=14 for blood samples), ii) non-tumor fraction parameters were set to c(0.95,0.99,0.995,0.999), iii) the ploidy was set to diploid, iv) the number of copy number states were reduced to 3, and v) subclonal copy number events were not considered. Samples with a tumor fraction below 5% or if no unequivocal SCNAs were detected, were classified as not informative (below the limit of detection, LOD).

To infer fragment length, raw sequencing data were first trimmed for bioinformatic analysis using the Cutadapt software.⁶⁰ After alignment to the human reference hg38 genome using bwa-mem, PCR duplicates were marked using MarkDuplicates⁶¹ and these were excluded from downstream analysis along with unmapped reads, reads of low mapping quality (<20), secondary and supplementary alignments. Fragment length was directly extracted from the alignment files. Plots were constructed in Python using the packages seaborn (version 0.13.2) and matplotlib (version 3.8.4).

Quantification and statistical analyses

Data collection and descriptive analyses were performed using Microsoft Excel (2016). Graphs and statistical analyses were done with GraphPad Prism (version 9.4.1.; GraphPad Software, San Diego, CA, USA). Wilcoxon matched-pairs signed rank and two-tailed Mann Whitney test were used for group comparisons. p < 0.05 was considered as statistically significant.

Published: September 23, 2025