Circulating tumour DNA as a promising biomarker for breast cancer diagnosis & treatment monitoring

(A) Diagnostics

ctDNA in real-time can detect cancer in asymptomatic patients, monitor disease progression, and evaluate treatment response in a non-invasive manner. In some cases, ctDNA can identify tumour mutations absent in tissue biopsy, providing a comprehensive genomic profile of the tumour²⁷. ctDNA detection strongly correlates with molecular subtypes, with the basal-like subtype exhibiting the highest ctDNA detection rate (86%), while luminal-A shows the lowest rate (0%)²⁸. Catarino et al²⁹ reported higher ctDNA levels in breast cancer patients than in healthy individuals (105.2 vs. 77.06 ng/ml). Furthermore, they demonstrated a notable reduction in ctDNA levels following surgery (59.0 vs. 105.2 ng/ml)²⁹. In another study, Zhou et al³⁰ reported that tumour-derived mutations were present in 85.71 per cent of patients with stage IV/M1 breast cancer, whereas only 57.81 per cent of stage I-III/M0 patients exhibited such mutations.

Although ctDNA is an effective biomarker in metastatic settings, ctDNA research for early breast cancer detection has yielded mixed results, as demonstrated by Kruspe et al³¹, showing uncertainties due to low CTC levels in early-stage cancer. Additionally, mutation rates and their specificity in ctDNA are relatively low, limiting its use in breast cancer screening. However, several instances exist where identification and analysis of specific gene mutations in cfDNA is used for breast cancer diagnosis. For instance, Arthrobacter luteus (ALU) sites that are widespread in the genome can be used to measure DNA integrity and are frequently utilised in breast cancer³². Using a fraction of ALU247 of total cfDNA levels instead of cfDNA alone led to a perfect 1.0 ROC score in differentiating between cancer-ridden and normal subjects; classifying patients with lymph node spread and those without revealed significant differences³³. Another study³⁴ correlates nodal spread using polymorphisms in specific regions of chromosome 13 and the PTEN gene using cfDNA. Methylation patterns in cfDNA have been successfully utilised for early detection using gene panels by several groups^35-37. Another notable study integrated ctDNA variant calling with plasma proteins profiling in the same blood sample, achieving >99 per cent specificity³⁸.

Detecting homologous repair deficient (HRD) tumours using BRCA variants obtained in cfDNA is a current frontier in diagnosis. Because cfDNA findings can give false negatives, no clinical guidelines specify liquid biopsy BRCA testing in lieu of conventional tests on FFPE sections³⁹. Secondly, the official FDA label of the PARP inhibitor olaparib has no statements indicating its application in breast cancer patients with somatic BRCA mutations, and ctDNA mutations cannot be confidently identified as somatic or germline^40,41. However, emerging reports demonstrate the occurrence of BRCA mutations in cfDNA of patients with HRD and compare somatic vs. germline BRCA mutation response to PARPi therapy, thus using liquid biopsies as a potential companion diagnostic^42-46. Additionally, epigenetic marks like promoter methylation of BRCA1, found by Harvey-Jones et al⁴⁷ in all repeat liquid biopsy samples taken from HRD tumours, demonstrate prognostic significance while circumventing the low sensitivity of ctDNA mutations.

ctDNA sequencing for cancer screening and early diagnosis remains challenging. Asymptomatic people only have ctDNA concentrations of 1 to 10 ng/ml³.Therefore, it was demonstrated that 150 to 300 ml of blood would be required for each test to reach 95 per cent sensitivity⁷. Second, healthy normal and haematogenic cells also supply the cfDNA pool, reducing specificity. The sensitivity and specificity standards for cancer screening and early diagnosis require considerable efforts and investments^48,49. Allele-specific real-time quantitative PCR (qPCR) techniques were primarily used in the early detection of ctDNA. The sensitivity/specificity limitations made for seldom application of technologies like Peptide Nucleic Acid (PNA) clamps, probe qPCR utilising TaqMan, and Scorpion Amplification Refractory Mutation System, in patients with significant tumour burden. However, this has changed over the last decade, with techniques like digital PCR (dPCR) and next-generation sequencing (NGS) having ultrasensitive detection thresholds for mutant allele abundance between 0.01 and 0.1 per cent.

The NGS edge in ctDNA diagnostics: Though sensitive and affordable, PCR-based techniques have low input and processing speed and can only screen for known variants. However, high throughput NGS can screen unidentified variations with a detection capability of mutant allele frequency (MAF) levels below one per cent. Other techniques, such as distinct molecular IDs or barcodes, can also improve sensitivity. With a MAF of 0.1 per cent, these techniques have identified 59 per cent of patients with stage I or stage II lung cancer with good concordance between the ctDNA response and radiographic response⁴⁹. Panel-based NGS can detect known variants in four ways: single locus, multiplex, targeted sequencing, and genome-wide sequencing. Several (including us) groups have recently developed NGS-based approaches that enable ultrasensitive ctDNA detection. Two of them used deep sequencing on a small set of amplicons that targeted frequently altered cancer genes. Although enabling low detection thresholds, the number of probable genomic locations has, to date, been constrained by technical issues with PCR assay multiplexing. Since most tumours won’t have a mutation identified by a specific tiny mix of amplicons, therapeutic applications are limited. The same bottleneck holds for translocations/rearrangements whose exact breakpoints aren’t known a priori⁵⁰.

Custom tumour-guided sequencing panel scans cover a range of variants. Some have examined up to 115 patient-specific mutations concurrently, measuring ctDNA to one mutant molecule every 33,333 copies in breast cancer patients after NACT. Such assays typically target 10-20 variants in plasma. Current methods can detect relapse in several cancer types, such as colorectal, breast, and bladder cancer, 3-10 months sooner than clinical relapse⁵¹. As whole-genome tumour sequencing is performed more often in clinical settings, there are fewer financial and logistical obstacles to implementing patient-specific panels⁵².

Of all sequencing methods, multigene targeted NGS panel is most suited for ctDNA sequencing of patients lacking tumour tissue, as it enables the identification of multiple variants in a one run of analysis. However, the presence of DNA non-shedding neoplastic lesions, artefacts produced during CHIP, or sequencing distort the sensitivity and specificity of NGS-based ctDNA testing⁵³. More research before clinical adoption is still required in this field.

(B) Minimal residual disease (MRD)

MRD (measurable, minimal or molecular residual disease) refers to residual tumour cells or biomarkers in the body post-treatment of cancer (both local and/or systemic). MRD might not exhibit any noticeable signs or symptoms and could be imperceptible with standard methods. Commonly employed techniques for MRD detection include qPCR, ddPCR, and NGS. Recent advancements in ctDNA analysis for identifying MRD in solid tumours post-curative therapy and before clinical or radiographic disease relapse have shown considerable therapeutic potential, as the detection of MRD through ctDNA analysis correlates with poor prognosis.

PCR is one of the basic techniques used to detect MRD, while BEAMing (beads, emulsion, amplification and magnetics) is a novel addition to the list. In a study by Beaver et al⁵⁴, ddPCR detected PIK3CA mutations in DNA from 30 breast cancer tumours and paired plasma samples before and after surgery. Liquid biopsy accurately identified MRD, finding 14 of the 15 PIK3CA mutations in pre-surgery ctDNA and detecting the same in about 50 per cent of patients post-surgery⁵⁴. However, both ddPCR and BEAMing require prior knowledge of specific mutations for detection, as they are designed to target predefined genetic changes. This restricts their ability to identify novel or unknown alterations, posing difficulties in cancers with significant heterogeneity or mutations that evolve during treatment^55,56. ddPCR has a MAF threshold of about 0.1 per cent, and BEAMing has one of 0.01 per cent, but rare mutations or heavily fragmented ctDNA are often missed^57,58. Additionally, they have limited multiplexing capability, a drawback when monitoring the clonal evolution of cancers such as leukaemia or lung cancer⁵⁹, making it time-consuming and technically demanding for routine clinical use^56,57. While ddPCR is relatively affordable, BEAMing requires specialised equipment and materials, limiting its adoption in resource-constrained settings⁵⁵. In contrast, NGS, especially targeted approaches like tagged-amplicons deep sequencing (TAm-Seq) and Safe Sequencing System (Safe-SeqS), overcomes some of these limitations by enabling broader and scalable mutation detection. However, NGS lacks sensitivity compared to ddPCR and BEAMing for detecting extremely low-frequency mutations, but a combined approach could offer a more robust solution for MRD detection^55,57,58.

Garcia-Murillas et al⁶⁰ demonstrated that targeted capture sequencing of the ctDNA can identify MRD-related genetic events, while primary tumour sequencing was less accurate at predicting future metastatic relapses. Therefore, mutation tracking can distinguish high versus low risks of relapse in newly diagnosed breast cancer patients. Issues of tumour heterogeneity could be partially overcome by subsequent adjuvant treatment therapies that are tailored to the genetic events occurring in the MRD. Janni et al⁶¹ used an NGS panel that included about 500 genes and around 4 Mb of epigenomic areas that experience differential methylation in several solid tumour types to demonstrate the viability of MRD detection using the Guardant Reveal assay.

Recently, McDonald et al⁶² developed TARgeted DIgital Sequencing (TARDIS), which uses exome sequencing of tumour biopsies to identify somatic mutations. Targeted linear pre-amplification of ctDNA with specific primers increases sensitivity. The test accurately detected ctDNA in all 33 patients tested--and also detected trace amounts in 17 out of 22 patients receiving NACT. Preliminary results in early-stage breast cancer suggest TARDIS can detect MRD with over 100 times the sensitivity of current methods⁶². Various studies that have demonstrated ctDNA as an independent indicator of the residual disease are described in table I^8,50,63,64.

C) Therapeutic resistance

ctDNA is increasingly used to detect therapeutic resistance in cancer treatment. As tumours evolve under treatment pressure, they may acquire genetic mutations that confer resistance to therapies, such as targeted treatments or chemotherapy.

Chen et al⁶⁵ identified resistance-correlated genes in liquid biopsies using NGS. Activating mutations in PIK3CA (C420R), TP53 and in HER2 (L869R) had correlations with HER2 targeted therapy resistance. Thompson et al⁶⁶ highlight the advantages of using ctDNA NGS to detect driver and resistance mutations, even when tissue samples are unavailable. While ctDNA testing alone cannot identify histologic sources of resistance, such as a small-cell phenotype, it can offer more detailed molecular characterisation of the tumour. The study also suggests that higher levels of cfDNA are linked to poorer survival outcomes, regardless of the mutational profile.

Unlike traditional imaging and serum biomarkers like CEA and CA15-3, which lack sensitivity for detecting minor tumour mutations, ctDNA provides a prompt and sensitive alternative for monitoring cancer progression. Specific ctDNA mutations can also guide treatment decisions, with examples including PIK3CA and BRCA1 mutations in MBC that suggest sensitivity to drugs like everolimus, alpelisib, and Olaparib, respectively. How ctDNA mutation patterns relate to hormone receptor/HER2 status and therapy is still unclear. A study by Davidson et al⁶⁷ showed that targeting AKT1 and ERBB2 mutations in ctDNA with capivasertib and neratinib, respectively, led to partial responses in 4/18 and 5/20 patients, suggesting the potential of mutation-driven therapies in late-stage, pre-treated breast cancer⁶⁴. Jacob et al⁶⁸ demonstrated that when comparing with tissue NGS, alterations detected in ctDNA at baseline showed a high degree of concordance. Notably, most of the tissue samples were taken from sites of metastatic disease, suggesting that resistance alterations may arise earlier in disease progression. Another explanation is that modest to moderate increases in clonal heterogeneity may not be captured due to dominant clones found in tissue samples.

A landmark study by Mosele et al⁶⁹ demonstrated the prevalence and clinical implications of PIK3CA mutations in different breast cancer subtypes. Among HR+/HER2- tumours, 28 per cent harboured a PIK3CA mutation, compared to 10 per cent in triple-negative breast cancer (TNBC). PIK3CA-mutated HR+/HER2- MBC showed reduced chemotherapy sensitivity and worse OS, whereas, in metastatic TNBC, PIK3CA mutation was associated with improved median OS⁶⁶. Continuously tracking treatment responses and ctDNA mutations in patients with MBC, there is potential to tackle drug resistance, though this requires confirmation through extensive prospective trials⁶⁰. In summary, ctDNA holds significant promise in oncology, particularly in advancing personalised medicine.

D) CtDNA as a prognostic marker for outcomes and response

Along with monitoring MRD and treatment resistance, ctDNA is also explored as a prognostic marker for survival and outcomes. Pècuchet et al⁷⁰ determined instead of being interpreted for its baseline concentration, the measurement of ctDNA in plasma throughout treatment should be evaluated independently. While baseline ctDNA measurements can provide initial insights into tumour burden and prognosis, they cannot capture tumour heterogeneity and predict treatment resistance. However, by serially measuring ctDNA levels over time, clinicians can detect early signs of recurrence, assess treatment response, identify emerging resistance mechanisms, and tailor treatment strategies^71,72. For instance, patients with initially low baseline ctDNA levels but a rapid increase during treatment had significantly worse outcomes compared to those with stable or declining levels⁷³. Thus, serial monitoring can identify high-risk patients who may need aggressive therapeutic interventions.

The absence of ctDNA normalisation in the initial assessment has a strong prognostic impact on both progression-free survival (PFS) and OS. As opposed to the MAF, which is a relative percentage statistic, the ctDNA concentration reported in studies is an absolute value of the MAF. Studies detected an inverse correlation between baseline ctDNA amount and the total amount of cfDNA and OS^74,75. Poorer outcomes are associated with ctDNA levels in early, locally advanced, and metastatic diseases.

Quality control for ctDNA

Blood is collected in BCTs, such as EDTA, Paxgene, and Strecktubes having proprietary elements that preserve the cfDNA and prevent cell lysis. Plasma samples are considered a better source of ctDNA as they provide wild-type DNA fractions. The overall cfDNA concentration in serum is 2-24 times higher, which is likely due to the contamination from DNA released from immune cells. Since clotting induces lysing of these immune cells, less contamination is found in plasma⁷⁶. Firstly, for extraction of cfDNA, blood is centrifuged at 1200×g for 10 min, then again at 16000×g for 10 min at 4°C. An overview of extraction and analysis of ctDNA from the resulting plasma layer has been illustrated in figure.

Close

Figure.

Extraction and analysis pipeline for circulating tumour DNA. The figure was created using CorelDraw 2024 (https://www.coreldraw.com/).

Export to PPT

Available kits for cfDNA extraction kits employ differing principles: (a) column-based separation, (b) spin column chromatography, and (c) size-based separation. Among these kits, column-based kits demonstrated better performance compared to the other two. This kit was the most effective, with a better representation of tiny-fragmented DNA. It also demonstrated maximum recovery of cfDNA and permitted an effective recovery of a spike-in. The basic column-based DNA extraction from whole blood kits was found to be more comparable to the column-based extraction kits by Repiská et al⁷⁷. However, a possible confounding effect is introduced by different adaptations of the DNA from whole blood kits for cfDNA extraction. An overview comparison of prevalent extraction methods is given in table II^78-80.

Extracted cfDNA is quantified using various kits and instruments as follows: (a) Qubit, (b) Tapestation, (c) Bioanalyzer, and (d) PCR. Fluorometric quantification, a streamlined approach compared to qPCR, has gained prominence in clinical laboratories due to its efficiency. By employing fluorescent probes selectively binding double-stranded DNA, fluorometry addresses the limitations inherent in spectrophotometry-based methods, such as NanoDrop. For downstream analyses requiring both quantification and cfDNA sizing, the concurrent utilisation of electrophoresis-based techniques, exemplified by TapeStation with various assays, is strongly recommended. Acceptable quality control value ranges for extracted ctDNA are described in table III^81-83. Furthermore, electrophoresis-based methods can be instrumental in assessing the quality of isolated cfDNA, particularly when pre-analytical factors, such as sample storage and shipping, may introduce the risk of gDNA contamination from cell lysis^81-83.

In summary, we note how ctDNA analysis complements conventional biopsy to provide a more representative sample of tumour mutations in a non-invasive manner. Its correlation with tumour burden makes it ideal for monitoring disease progress, especially in the MRD setting. The main innovations required before ctDNA tests become a routine diagnostic tool include better methods of identifying the ctDNA population in the liquid biopsy, research into novel prognostic and actionable data informed by ctDNA analysis specific to cancer subtypes and lowering costs of sequencing. As with all novel techniques, awareness of the transforming impacts of ctDNA testing among clinicians remains the primary factor deciding its adoption in routine practice.