Liquid biopsy-based multi-cancer early detection (MCED) assays are poised to revolutionize cancer diagnosis, providing prompt intervention and potentially reducing cancer-related mortality and healthcare costs1. Several MCEDs are based on the analysis of plasma cell-free DNA (cfDNA)1,2. In healthy individuals, cfDNA fragments in plasma are mainly derived from the hematopoietic lineage, while in cancer patients a fraction is derived from tumor tissue (circulating tumor DNA, ctDNA)3. The acquired (epi)genetic characteristics of the tumor DNA are reflected in the ctDNA fragments, including single-nucleotide mutations, copy-number alterations (CNA), methylation patterns, and fragmentomic features such as nucleosome positioning or end motives4. These features have been leveraged for the development of several MCEDs2,4.

Despite the promise of cfDNA-based MCED technologies, several challenges limit their clinical implementation. Test positivity is not always caused by the presence of a malignant condition, but can also be due to indolent diseases, premalignant conditions, or benign lesions5,6. Moreover, abnormalities of uncertain significance (AUS) can be detected, for which no explanation can be offered. As a result, screening assays could lead to anxiety, overdiagnosis and overtreatment, which could negatively impact the participants in MCED screening programs5,7. A major limitation is the lack of knowledge about the evolution and clinical consequences of these abnormalities of unknown significance. Systematic long-term follow-up of such individuals would address this knowledge gap.

During a genome-wide plasma cfDNA CNA-based cancer screening study conducted on 1002 elderly volunteers without cancer history8, we previously reported aberrant cfDNA CNA profiles in 24 participants for whom downstream clinical investigations did not identify a cancer. For 9 (37.5%), the CNAs (i.e., del(3p-), del(5q-), del(20q-), del(22q-) and trisomy 15) were shown to originate from mosaic copy-number alterations (mCA) in peripheral blood cells8. The origin of the CNAs in the remaining 15 participants could not be traced to blood. To understand the clinical relevance of those CNAs of unknown significance, we conducted a follow-up analysis 3–5 years later to assess changes in plasma cfDNA.

We contacted the 15 participants with CNAs of unknown significance in the initial CNA-based cancer screening study (n = 15, mean age at follow-up investigations: 74.5 years; range: 69–84 years; male/female: 11/4). Two participants died in the follow-up period. The remaining 13 participants were sent a self-reporting questionnaire, and 12 responded (average time elapsed: 4.1 years; range: 3.4–5.1 years; Supplementary Fig. 1). None of the responders reported a cancer diagnosis.

We next assessed whether and how the CNAs of unknown significance in plasma cfDNA evolved. Of the 12 participants who responded to the questionnaire, 9 donated a new plasma sample (Supplementary Fig. 1). In two participants (cases 17112016-02 and 06022017-06), the CNAs of unknown significance in the plasma cfDNA profile disappeared (Table 1; Supplementary Fig. 2). In 5 individuals, the original single chromosomal aneuploidy remained similar (Table 1; Fig. 1; Supplementary Fig. 3). In two, the cfDNA CNA profiles were more outspoken and expanded. In addition to the initial genome-wide gains and losses of multiple chromosomes, there was a trisomy 15 in case 02022017-08. In case 29042016-27, the initial segmental del(5q), progressed into a profile with several aneuploidies (Table 1, Fig. 1).

Table 1 Evolution in time of chromosomal aberrations in cell-free and genomic DNA from WBC, as measured by GIPSeq and FISH, respectively
Full size table
Fig. 1: Visualization of the evolution of CNAs in cell-free plasma DNA and peripheral blood cells at the time of the initial CNA cancer screening study and the follow-up.
figure 1

A, B Representative CIRCOS plots showing the evolution of the CNA detected in cfDNA with GIPseq results from the initial CNA cancer screening study (outer circle) versus follow-up (inner circle). Each plot represents signals with a z-score ≥3.0 (suggesting gain; in green) or ≤3.0 (suggesting loss; in red). The chromosomes with a |zz-score| ≥ 3.0 are marked with a blue rectangle, where the score is calculated as the standard score of the z-score of an autosome compared to the z-score of the remaining ones. C, D FISH analysis on freshly isolated peripheral blood cells of the observed chromosomal aberrations. A genomic representation profile and C dual-color FISH of evolved cases at the follow-up, 02022017-08 and 29042016-27. B genomic representation profiles and D dual-color FISH of stable cases at the follow-up, 18022016-01 and 20042017-10. C, D, left panels: three signals for 15q (PML, Spectrum Orange) and two signals for 17q (RARA, Spectrum Green). C right panel: one signal for chromosome 5q (EGR1, Spectrum Orange) and two signals for 5p (D5S23, Spectrum Green). D right panel: loss of 9p (Spectrum Orange), two signals for CEP 12 (Spectrum Aqua).

Full size image

All seven subjects with a persistent CNA profile in their plasma were invited to undergo clinical investigations. Five participants’ general blood parameters (i.e., cell count and cytology) were within normal ranges. Two (20042017-10 and 02022017-08) presented with anemia, yet this finding was likely unrelated to the detected CNA in cfDNA. For none of the cases, WB-DWI MRI screening indicated a malignancy. FISH using probes located within the identified CNAs demonstrated the presence of low-grade mosaicism in the white blood cells for trisomy 15 in three cases (3–4% positive cells; Table 1; Fig. 1; Supplementary Fig. 3), showed a normal diploid constitution in one and demonstrated medium-grade mosaic aneuploidy in the peripheral blood of two cases (Table 1; Fig. 1). FISH failed in one case.

In conclusion, for 5/9 individuals with CNA of unknown significance during the initial cfDNA cancer screening, (del(9q), del(5q) or trisomy 15), became detectable by FISH in peripheral blood cells 3–5 years later.

This study shows that AUSs can disappear or remain stable, being of hematological origin. We hypothesize that some AUSs are transient. The presence of clones with CNAs in blood cell populations might be negatively selected. For example, the presence of a CNA can lead to abnormal protein expression, allowing the recognition of the carrying clone by the immune cells (e.g., via major histocompatibility complex molecules), leading to immune-mediated negative selection9,10.

The CNAs detected in plasma cfDNA persisted in seven cases. Whereas the origin of the CNAs could not be determined in the first study, we demonstrate their presence in white blood cells in five cases, illustrating the sensitivity of cfDNA CNA analysis can be higher than traditional methods to detect clonal hematopoiesis. The aberrations in the two cases with medium-grade mosaicism (del(5q) and del(9q)) are both known to be associated with myelodysplastic syndromes (MDS), acute myelogenous leukemia and myeloproliferative neoplasms11,12. However, neither of these subjects showed clinical signs of these disorders. Three cases with isolated trisomy 15 in cfDNA showed low-grade mosaicism (3–4%) in peripheral blood cells. Clonal mosaicism for trisomy 15 in peripheral blood cells has previously been reported in the elderly, with a higher male prevalence13. However, it has rarely been reported as a sole abnormality in hematological disorders13.

Clonal haematopoiesis14 defines a premalignant state predominantly found in elders that can confound the interpretation of MCEDs6,8,15. This age-related phenomenon is associated with a 10-fold increased risk of developing a hematological malignancy14,16, although there is a lack of knowledge on which alterations effectively lead to a malignant transformation or other clinical manifestation (e.g., alteration in the blood parameters or cardiovascular problems)14. Molecular alterations in CH are most commonly single-nucleotide variants (SNVs), referred to as clonal hematopoiesis of indeterminate potential (CHIP), with an incidence of approximately 10-30% in older adults16. In contrast, the presence of large copy-number alterations, defined as mosaic chromosomal alterations (mCA), has a lower incidence, estimated at around 1–5%17. While the confounding role of CHIP in the interpretation of plasma cfDNA-based cancer analysis is well-established15,18, the role of mCAs in blood cells has only recently been recognized8,19. Along with the knowledge of the possible confounding effect of this phenomenon on the interpretation of MCEDs test results, we demonstrate here that signals of unknown significance discovered in plasma cfDNA can be of hematopoietic origin.

The findings could have clinical care consequences and be translated into guidelines. When bringing MCEDs assays into the clinical context, a thorough consideration of the possible clinical significance of the detected aberrations is crucial to determine the next steps1,7. In fact, both transient signals and clonal hematopoiesis, particularly in the form of mCAs, should be considered during test results interpretation. Given the uncertainty on the individual risk of malignant transformation of clonal hematopoiesis and the length of the possible lagging time between detection and tumor clinical manifestation16, it is important to inform patients carefully to avoid overinterpreting the screening tests.

Both in our study as well as in clinical guidelines for the detection of indolent cancers in pregnancy20, patients undergo different clinical examinations to identify the origins of the CNAs, including complete blood count (CBC) and WB-MRI. Similarly, the latest hematological guidelines indicate that monitoring of the patient with CH should be individualized21. For individuals under surveillance, current guidelines generally recommend follow-up every three to six months consisting of interval history, physical examination, and laboratory testing (e.g., CBC with differential)21. Regular repetition of DNA sequencing or bone marrow examination is not recommended unless there is an indication (e.g., clinical change). This approach helps avoid unnecessary procedures and focuses on relevant diagnostic steps. Additionally, the integration of bioinformatic pipelines to filter the non-cancer-related calls based on databases would further enhance the accuracy of the MCED tests. For instance, the use of cfDNA deconvolution methods could help distinguish the tissue of origin of the alterations to further guide clinical decisions22. Establishing guidelines describing an adequate follow-up window and defining how often and in which cases longitudinal cfDNA analysis or further genetic and/or clinical investigation are needed is of utmost importance before bringing these liquid biopsy-based methods into the routine clinical application. Nevertheless, further investigations into the underlying causes of signals with uncertain significance in the context of cfDNA-based MCED assays are needed to map the full scale of potential clinical outcomes and enable the establishment of evidence-based guidelines.

Methods

Study design and participants

Participants were elderly volunteers who participated in a former cfDNA CNA cancer screening study executed between October 2015 and June 20178. Between 01/02/2021 and 20/02/2021, participants with CNAs of unknown significance were recontacted and invited to (i) complete a health questionnaire, (ii) undergo a clinical examination, and (iii) donate a new plasma cfDNA sample (Supplementary Fig. 1). The study was approved by the ethics committee of University Hospitals Leuven (S63983) in accordance with the declaration of Helsinki. Written informed consent was obtained from all recontacted participants.

Self-reporting health questionnaire

In a self-reporting health questionnaire, participants were asked (i) whether an interim cancer diagnosis had been made and, if positive, (ii) which cancer type was diagnosed and (iii) when the diagnosis was made. When a cancer diagnosis was reported, the patient’s medical file was consulted to retrieve detailed information about the cancer type and staging. In case the participant was deceased before the questionnaire could be answered, the general practitioner was contacted to ascertain the cause of death, and, in case of a cancer diagnosis, details were obtained about the tumor characteristics.

Clinical investigations

The participants were invited to undergo whole-body diffusion-weighted MRI (WB-DWI MRI) imaging to screen for cancerous lesions, similar to the initial CNA cancer screening study8. In parallel, a peripheral blood sample was analyzed for general hematological parameters (cell counts and cytology), clinical biochemistry and protein electrophoresis. This was done at the Laboratory Medicine of University Hospitals Leuven following standard procedures, as described before8.

Low-pass whole-genome sequencing of cfDNA and data analysis

A prospective peripheral blood sample was taken for cfDNA CNA analysis (cfDNA blood collection tubes®, Roche, Switzerland). Plasma cfDNA extraction, low-pass whole-genome sequencing of cfDNA and subsequent genome-wide detection of CNAs were performed using the in-house GIPseq pipeline, as in the initial CNA-based cancer screening study8. Briefly, chromosome-dependent parameters (z-score and zz-scores), and a genome-wide quality score (QS) were used to classify a GIPSeq profile as suspicious of cancer23. For each participant, the GIPseq profile in the novel blood sample was compared to the baseline result when participating in the initial CNA-based cancer screening study. A GIPSeq profile was ranked as “normalized” if the follow-up profile had a QS < 2 and there were no significant gains or losses anymore across one of the chromosomes (i.e., all |z-score | < 3.0 and |zz-score | < 3.0). The novel GIPSeq profile was marked as “status quo” when showing a CNA profile similar to that of the baseline sample. A GIPSeq profile was marked as “evolved” if additional aberrant CNAs were found in the novel sample (i.e., |QS | > 2 or additional chromosomes with a |z-score | > 3.0 and |zz-score | > 3.0). The limit of detection of GIPSeq was previously established23. The method can detect a trisomy present in as low as 3% of cells with over 99% accuracy when sequencing at a depth of 10 million reads.

Fluorescent in situ hybridization

Fluorescent in situ hybridization (FISH) was carried out on peripheral blood cells according to standard procedures. The DNA probes are described in Supplementary Table 1.