Longitudinal profiling of hormone receptor positive, HER2 negative metastatic breast cancer through droplet digital PCR-based circulating tumor DNA fragmentomics

Highlights

• •ACTB fragmentomics represents a reliable surrogate for tumor dynamics in first-line HR+ HER2- MBC
• •High ACTB_short levels negatively impact both progression-free survival and overall survival.
• •ACTB_medium fragments show a favorable impact on progression-free survival.
• •ACTB fragmentomics may represent a mutation-agnostic, early risk stratification tool for MBC monitoring.

Abstract

Background

In the context of hormone receptor positive, HER2 negative Metastatic breast cancer (MBC), CDK 4/6 inhibitors (CDK4/6i) combined with endocrine therapy represent the standard first-line treatment, improving Progression-Free Survival (PFS) and Overall Survival (OS). Despite these benefits, resistance to treatment develops, necessitating early risk classification to guide clinical management. This study explores the potential of cell-free DNA (cfDNA) fragmentomics, specifically ACTB fragments, in predicting tumor dynamics and treatment outcomes in luminal MBC, based on the principle that shorter DNA fragments are generally indicative of circulating tumor DNA (ctDNA) from tumor cells, while longer fragments are associated with leukocyte lysis.

Methods

In the MAGNETIC.1 study, 141 women with luminal-like MBC were enrolled between January 2018 and January 2023. Blood samples were collected at baseline (BL), and after 3 (T3) and 6 (T6) months of treatment. cfDNA was extracted and analyzed using droplet digital PCR (ddPCR) to quantify ACTB fragments (136 bp, 420 bp, and 2,000 bp). Continuous variables were compared using the Mann–Whitney test and Kruskall Wallis test depending on data distribution and number of groups. Categorical variables were compared using the Chi-square test or Fischer’s exact test whenever appropriate. Differences in survival were tested by log-rank test and uni- and multivariable Cox regression.

Results

By categorizing the values of actinic fragments into interquartiles (Q1, Q2, and Q3), ACTB_short Q3 at baseline was significantly associated with negative PR expression (RRR 0.27, P = 0.012) and a higher frequency of liver metastasis (RRR = 3.75, P = 0.009). In terms of clinical outcomes, regarding PFS a significant role was observed for baseline ACTB_short Q3 (HR 1.92, P = 0.041) and ACTB_medium Q3 (HR 0.47, P = 0.043), the latter maintaining significance in multivariable analysis (HR 0.33, 95 %, P = 0.012). For OS, ACTB_short Q3 demonstrated a significant impact in both univariable (HR 3.94, P = 0.003) and multivariable analyses (HR 3.25, P = 0.023).

Conclusions

This study demonstrates the feasibility of employing a fragmentomics mutation agnostic approach in luminal MBC. Baseline and longitudinal changes in ACTB fragments were significantly associated with clinical outcomes, suggesting their potential as non-invasive biomarkers for early risk classification and monitoring tumor dynamics.

Background

Metastatic breast cancer is the leading cause of cancer-related deaths among women, resulting in over 600,000 fatalities annually [1]. The luminal subtype accounts for more than 70 % of MBC diagnoses. In this context, the combination of CDK 4/6 inhibitors and endocrine therapy as a first-line treatment has been shown to significantly improve PFS and, in some cases, overall survival OS [2]. However, nearly all patients eventually develop resistance to treatment, with 10–30 % developing primary resistance, emphasizing the importance of early risk classification to guide clinical management [3,4].

Extensive efforts have been dedicated to understanding the mechanisms underlying this resistance, aiming to predict tumor dynamics and guide treatment choices in subsequent lines. Liquid biopsy is increasingly emerging as a crucial non-invasive tool for assessing tumor biology both statically and longitudinally, as well as for evaluating treatment responses. To date, the analysis of circulating tumor DNA (ctDNA) and Circulating Tumor Cells (CTCs) are the primary parameters explored for real-time early cancer detection, monitoring minimal residual disease, and longitudinally tracking clonal evolution in peripheral blood [5]. However, the challenge of identifying mutations that are not always detectable in plasma for ctDNA analysis, combined with the difficulties of analyzing CTCs due to their low abundance and associated costs, continues to make these approaches difficult to implement in all patients.

In this context, the study of cell-free DNA fragmentation patterns, known as fragmentomics, is progressively recognized as a crucial, mutation-agnostic tool for the longitudinal analysis of tumor evolution [6]. Our previous work has utilized a fragmentomics approach to assess the proportion of ctDNA by analyzing ACTB DNA fragments, categorized into lengths of 136 bp, 420 bp, and 2,000 bp and termed ACTB_short, ACTB_medium, and ACTB_long, respectively [7]. The underlying principle of this approach is that plasma DNA stems from distinct biological processes: short fragments are typically associated with ctDNA and reflect fragmentation patterns characteristic of tumor cell release, whereas medium and long fragments are generally attributed to genomic DNA (gDNA) released from non-tumoral cells, such as leukocytes, through cellular lysis [8,9]. This fragment-based stratification may therefore provide indirect yet informative insights into tumor burden and its dynamic changes during treatment.

In this study, we validated our ddPCR fragmentomics pathway and further explored it for the longitudinal profiling of first-line luminal MBC. By leveraging a more mature dataset, with a larger patient cohort and an extended follow-up period, this analysis enables a more comprehensive examination of the distribution and dynamics of ACTB fragments, uncovering their biological significance and their potential prognostic value in both PFS and overall survival OS.

Materials and methods

Study population and ethics statement

A cohort of 141 women was prospectively enrolled in the MAGNETIC.1 (NCT05814224) multicenter pragmatic study, between January 2018 and January 2023. All patients were diagnosed with luminal-like MBC and received either fulvestrant or aromatase inhibitors (Ais) with or without CDK4/6i as first-line ET according to the investigator’s choice. The two main exclusion criteria were a diagnosis of any secondary malignancy within the last three years and prior ET for MBC. Patients could have received both ET and chemotherapy in the adjuvant and neoadjuvant setting. Samples of 134 patients were collected before treatment start (BL) and after 3 and 6 months concomitantly with computed tomography (CT) scan restaging [first (T3) and second (T6) evaluation, respectively]. The study was approved by the ethics committee under the CEUR-2018-Sper-056-CRO protocol.

Extraction of circulating tumor DNA from plasma samples

Blood samples were collected using the PAXgene Blood ccfDNA Tubes (Qiagen) or the Cell-Free DNA BCT tubes (Streck). ctDNA was isolated from 4.8 ml of plasma with the QIAsymphony PAXgene Blood ccfDNA Kit (PreAnalytiX) through the QIAsymphony SP instrument (Qiagen) using the recommended Standard Protocol Line (STA) for small fragment enrichment and eluted in 60 μl of elution buffer (Qiagen). ctDNA concentration was estimated using the Qubit 1X dsDNA HS Assay Kit (Qiagen).

Droplet digital PCR

Circulating free DNA (cfDNA) samples were analyzed for cytolysis-derived contamination and ctDNA quantification by ddPCR though the detection of small (136 bp), medium (420 bp) and long (2000 bp) ACTB fragments. The presence of short ACTB is associated with ctDNA, while the medium and long fragments are associated with gDNA [5]. Based on this biological rationale, 5’ 6-FAM- conjugated short actin and 5’ HEX-conjugated medium actin assay (Bio-Rad) were used. Samples were run in duplicate and then merged for further analysis. Droplets were generated on the automated droplet generator QX200 AutoDG™ (Bio-Rad) and after PCR amplification, droplets were read on the QX200™ Droplet Reader (Bio-Rad). Data were analyzed using the QuantaSoft™ 1.7.4 Software (Bio-Rad). Based on FAM and HEX probe fluorophores, short actin is Ch1+Ch2- (FAM), medium actin is Ch1-Ch2+ (HEX) and long actin is Ch1+Ch2+ (positive for FAM and HEX) were considered for analysis.

Statistical analysis

Clinical and pathologic variables were reported using descriptive analyses. Categorical variables were reported as frequency distributions, whereas continuous variables were described through median and interquartile ranges (Qs). Continuous variables were compared using the Mann–Whitney test and Kruskall Wallis test depending on data distribution and the number of groups. Categorical variables were compared using the Chi-square test or Fischer’s exact test, as appropriate.

Variations over time of short actin fragments, medium actin fragments and long actin fragments were evaluated with generalized linear mixed-effects models (LME) for repeated measures. Multiple comparisons of parameters’ values across different follow-up periods (baseline, 3 months, and 6 months) were performed using the Friedman test for repeated measures, with p-values adjusted using the Bonferroni post-hoc test. A multivariate multinomial logistic regression model was adopted to assess the association between clinical factors (independent variables) and actin fragments (dependent variable, coded as, Q1,Q2 and Q3 considering Q2 as the reference category). The regression coefficients are expressed as relative risk ratio (RRR).

PFS was defined as the time from BL to progression (as determined by imaging) or death for any cause, whichever occurred first. OS was defined as the time from BL until death from any cause. Patients without an end point event at the last follow-up visit were censored. The statistical significance of associations between individual variables and survival was calculated using the log-rank test. Results are expressed as hazard ratios (HRs) and 95 % confidence intervals (95 % CIs).

Statistical analysis was conducted using StataCorp 2016 Stata Statistical Software: Release 15.1 (College Station, TX), R (version 3.3.1; The R foundation for Statistical Computing, Vienna, Austria) and JMP (version 14; SAS Institute, Cary, NC).

Results

A cohort of 134 patients with luminal-like MBC treated with first-line ET was prospectively enrolled between May 2018 and January 2023. Of the 134 patients, 24 % had progesterone receptor (PR)-negative MBC, while the proportion of HER2-negative and HER2-low disease was 35 % and 65 %, respectively. Bone was the most common metastatic site (70 %), followed by lymph nodes (54 %) and liver (28 %). Sixty-two percent of patients had a de novo diagnosis of MBC. The first-line ET regimen was mainly based on AIs (72 %) compared to fulvestrant, which was used in 28 % of cases. The proportion of patients who received a CDK 4/6 inhibitor was 96.26 %, with palbociclib being the most frequently administered (47.29 %), followed by ribociclib (38.76 %) and abemaciclib (13.95 %) (Table 1).

Table 1.

Cohort characteristics; missing data were not reported.

	N	%	ACTBshort			Chi2	ACTBmedium			Chi2	ACTBlong			Chi2
			0–25	25–75	>75		0–25	25–75	>75		0–25	25–75	>75
Histotype
IDC	101	83.47	25 (26.04 %)	49 (51.04 %)	22 (22.92 %)	0.515	24 (25 %)	48 (50 %)	24 (25 %)	0.310	25 (26.04 %)	47 (48.96 %)	24 (25 %)	0.566
ILC	20	16.53	6 (31.58 %)	7 (36.84 %)	6 (31.58 %)		4 (21.05 %)	7 (36.84 %)	8 (42.11 %)		4 (21.05 %)	8 (42.11 %)	7 (36.84 %)
PR
Negative	31	23.48	6 (20 %)	11 (36.67 %)	12 (43.33 %)	0.035	6 (20 %)	15 (50 %)	9 (30 %)	0.891	3 (10 %)	16 (53.33 %)	11 (36.67 %)	0.097
Positive	101	76.52	25 (26.04 %)	52 (54.17 %)	19 (19.79 %)		23 (23.02 %)	47 (48.96 %)	26 (27.08 %)		27 (28.12 %)	46 (47.92 %)	23 (23.96 %)
Lung
Yes	32	24.06	5 (16.67 %)	19 (63.33 %)	6 (20 %)	0.218	9 (30 %)	14 (46.67 %)	7 (23.33 %)	0.616	6 (20 %)	16 (53.33 %)	8 (26.67 %)	0.788
No	101	75.94	27 (27.84 %)	44 (45.36 %)	26 (26.80 %)		21 (21.65 %)	48 (49.48 %)	28 (28.87 %)		25 (25.77 %)	46 (47.42 %)	26 (26.80 %)
Liver
Yes	37	27.82	11 (29.73 %)	12 (32.43 %)	14 (37.84 %)	0.033	11 (29.73 %)	19 (51.35 %)	7 (18.92 %)	0.315	12 (32.43 %)	14 (37.84 %)	11 (29.73 %)	0.242
No	96	72.18	21 (23.33 %)	51 (56.67 %)	18 (20 %)		19 (21.11 %)	43 (47.78 %)	28 (31.11 %)		19 (21.11 %)	48 (53.33 %)	23 (25.56 %)
Bone
Yes	93	69.92	24 (26.97 %)	41 (46.07 %)	24 (26.97 %)	0.475	23 (25.84 %)	39 (43.82 %)	27 (30.34 %)	0.226	24 (26.97 %)	41 (46.07 %)	24 (26.97 %)	0.532
No	40	30.08	8 (21.05 %)	22 (57.89 %)	8 (21.05 %)		7 (18.42 %)	23 (60.53 %)	8 (21.05 %)		7 (18.42 %)	21 (55.26 %)	10 (26.32 %)
Node
Yes	72	54.14	13 (18.57 %)	37 (52.86 %)	20 (28.57 %)	0.153	17 (29.82 %)	25 (43.86 %)	15 (26.32 %)	0.323	13 (18.57 %)	35 (50 %)	22 (31.43 %)	0.175
No	61	45.86	19 (33.33 %)	26 (45.61 %)	12 (21.05 %)		17 (29.82 %)	25 (43.86 %)	15 (26.32 %)		18 (31.58 %)	27 (47.37 %)	12 (21.05 %)
CNS
Yes	2	1.50	0	2 (100 %)	0	0.356	0	1 (50 %)	1(50 %)	0.653	0	2 (100 %)	0	0.345
No	131	98.50	32 (25.60 %)	61 (48.80 %)	32 (25.60 %)		30 (24 %)	61 (48.80 %)	34 (27.20 %)		31 (24.80 %)	60 (48 %)	34 (27.20 %)
De Novo
Yes	60	61.86	7 (19.44 %)	20 (55.66 %)	9 (25 %)	0.280	3 (8.33 %)	20 (55.56 %)	13 (36.11 %)	0.163	5 (13.89 %)	20 (55.56 %)	11 (30.56 %)	0.301
No	37	38.14	20 (33.90 %)	29 (49.15 %)	10 (16.95 %)		14 (23.73 %)	28 (47.46 %)	17 (28.81 %)		16 (27.12 %)	26 (44.07 %)	17 (28.81 %)
Type of CDK
Palbociclib	61	47.29	21 (35.59 %)	32 (54.24 %)	6 (10.17 %)	<0.001	15 (25.42 %)	35 (59.32 %)	9 (15.25 %)	0.035	14 (23.73 %)	38 (64.41 %)	7 (11.86 %)	<0.001
Ribociclib	50	38.76	5 (10.42 %)	25 (52.08 %)	18 (37.50 %)		9 (18.75 %)	19 (39.58 %)	20 (41.67 %)		8 (16.67 %)	18 (37.50 %)	22 (45.83 %)
Abemaciclib	18	13.95	5 (31.25 %)	25 (31.25 %)	6 (37.50 %)		5 (31.25 %)	6 (37.50 %)	5 (31.25 %)		9 (56.25 %)	3 (18.75 %)	4 (25 %)
Endocrine companion
Fulvestrant	38	28.15	11 (30.56 %)	16 (44.44 %)	9 (25 %)	0.629	11 (30.56 %)	18 (50 %)	7 (19.44 %)	0.368	11 (30.56 %)	17 (47.22 %)	8 (22.22 %)	0.542
AI	97	71.85	21 (22.83 %)	48 (52.17 %)	23 (25 %)		20 (21.74 %)	44 (47.83 %)	28 (30.43 %)		20 (21.74 %)	46 (50 %)	26 (28.26 %)

ACTB segments are associated with baseline clinical characteristics

At baseline the median levels were 37 copies for ACTB_short (IQR 18.5–110), 2 (Q 1–6) for ACTB_medium and 14.2 (IQR 7.5–32) for ACTB_long (Supplementary Table 1). After 3 months (T3), median levels were 28 (IQR 11.5–60.5) for ACTB_short, 1 (IQR 0.5–4.5) for ACTB_medium, and 10 (IQR 4.5–19) for ACTB_long (Supplementary Table 1). At 6 months (T6), the median levels were 58.25 (IQR 36–87) for ACTB_short, 1 (IQR 0–2.5) for ACTB_medium, and 12.75 (IQR 7–20.5) for ACTB_long (Supplementary Table 1). Significant differences were observed among ACTB_short, ACTB_medium, and ACTB_long in the general population at baseline (P < 0.001), at T3 (P < 0.001), and T6 (P < 0.001) (Fig. 1).

Fig. 1.

ACTB_long(long) ACTB_short (short), ACTB_medium (medium) fragments distribution across the three investigated timepoints (T0, T3, T6). Median, interquartile range, and outliers are described for the overall biomarker distribution at each timepoint through box plots. Outliers were removed. Abbreviations: T0, baseline; T3, after 3 months; T6, after 6 months.

By categorizing the values of actinic fragments into interquartiles (0–25, 25–75, 75–100, respectively Q1, Q2, and Q3) and assessing their distribution in relation to clinical variables, a significant difference was observed for ACTB_short concerning PR status (P = 0.035) and liver metastases (P=0.033), and for ACTB_short, ACTB_medium, and ACTB_long with regards to the type of CDKi (P<0.001, P = 0.035, P < 0.001, respectively) (Table 1).

The association between clinical and molecular features across actin fragments Q was further explored through uni- and multi-variable multinomial logistic regression analyses. In the univariable analysis, concerning the intermediate group (Q2), ACTB_short Q3 was significantly associated with negative PR expression (Relative Risk Ratio [RRR]=0.31, P = 0.017) and a higher risk of liver metastasis (RRR 3.31, P = 0.013) (Supplementary Table 2). This association was also confirmed in the multivariable model (respectively RRR 0.27, P = 0.012 and RRR = 3.75, P = 0.009) (Table 2). No significant association was observed in relation to Actin_medium and Actin_long fragments (Supplementary Tables 3 and 4).

Table 2.

Multivariable multinomial logistic regression for Actin_short Q1 and Actin_short Q3. Actin_short Q2 was considered as base subgroup. Abbreviations: PR=Progesteron Receptor, Q= Interquartile Range.

	Relative Risk Ratio	95 % Confidence interval		P
Q1
PR	0.84	0.28	2.54	0.754
Liver metastases	2.04	0.76	5.47	0.154
Q2
PR	1.00 (Ref)
Liver metastases	1.00 (Ref)
Q3
PR	0.28	0.10	0.75	0.012
Liver metastases	3.65	1.38	9.69	0.009

ACTB dynamics across fragments subgroups after 3 and 6 months of treatment

We then analyzed the dynamics of actinic fragments across three timepoints. Median levels of ACTB_short significantly decreased between T3 and T0 (P < 0.001) and significantly increased between T6 and T3 (P < 0.001), while the reduction between T6 and T0 was not significant (P = 0.621) (Supplementary Fig. 1).

Median levels of ACTB_medium significantly decreased between T3 and T0 (P = 0.003) and T6 and T0 (P < 0.001), while the reduction between T6 and T3 was not significant (P = 0.220) (Supplementary Fig. 2).

Median levels of ACTB_long significantly decreased between T3 and T0 (P < 0.001), showed a non-significant increase between T6 and T3 (P = 0.645) and a non-significant reduction between T6 and T0 (P = 0.059) (Supplementary Fig. 3).

Sites of disease did not impact on the dynamics of ACTB_short, ACTB_medium and ACTB_long (Supplementary Tables 5–and 7).

Baseline levels of ACTB_short and ACTB_medium significantly impact on PFS and OS

The prognostic impact of ACTB_short, ACTB_medium, and ACTB_long, was then investigated in terms of PFS and OS after interquartile categorization, using the intermediate interquartile as reference.

At baseline, across clinical variables, a significant impact on PFS emerged for lobular histotype (Hazard Ratio [HR] 2.23, 95 % Confidence Interval [CI] 1.19–4.17, P = 0.012) and the presence of liver metastases (HR 2.93, CI 1.79–4.78, P<0.001) (Table 3). Regarding ACTB fragments, an inverse significant role was observed for ACTB_short Q3 (HR 1.92, 95 % CI 1.03–3.59, P = 0.041) and ACTB_medium Q3 (HR 0.47, 95 % CI 0.23–0.97, P = 0.043) (Fig. 2, Table 3). ACTB_short Q1vs Q2 was associated with worse PFS but did not reach statistical significance (HR 1.67, CI 0.95–2.93 P = 0.077) (Fig. 2, Table 3). In multivariable analyses a significant impact on PFS was confirmed for lobular histotype (HR 3.04, CI 1.56–5.93, P = 0.001), liver metastases (HR 2.29, CI 1.33–3.97, P = 0.003) and ACTB_medium Q3 (HR 0.33, 95 % CI 0.12–0.76, P = 0.012) (Table 3). PFS at 12, 24 and 36 months for the respective interquartiles of ACTB_short, ACTB_medium and ACTB_long are reported in Table 5.

Table 3.

Univariate and Multivariate Cox regression analysis for Progression-Free Survival (PFS).

	Univariate Analysis			Multivariate Analysis
	HR	95 % CI	p-value	HR	95 % CI	p-value
PR status
Negative	1.00 (Ref)
Positive	0.61	0.35–1.06	0.081
De novo MBC
Yes	1.00 (Ref)
No	0.61	0.33–1.11	0.108
Histotype
Ductal	1.00 (Ref)			1.00 (Ref)
Lobular	2.23	1.19- 4.17	0.012	3.04	1.56–5.93	0.001
Liver Metastasis
No				1.00 (Ref)
Yes	2.93	1.79–4.78	<0.001	2.29	1.33–3.97	0.003
Lung Metastasis
No	1.00 (Ref)
Yes	0.57	0.30–1.08	0.085
CNS Metastasis
No	not evaluable
Yes
Bone Metastasis
No	1.00 (Ref)
Yes	1.27	0.73–2.20	0.399
Node Metastasis
No	1.00 (Ref)
Yes	1.17	0.72–1.90	0.526
ACTBshort
Q2	1.00 (Ref)			1.00 (Ref)
Q1	1.67	0.94–2.93	0.077	1.17	0.63–2.17	0.622
Q3	1.92	1.03–3.59	0.041	1.97	0.93–4.19	0.079
ACTBmedium
Q2	1.00 (Ref)			1.00 (Ref)
Q1	1.04	0.60–1.80	0.887	1.05	0.59–1.89	0.863
Q3	0.46	0.23–0.97	0.043	0.33	0.14–0.78	0.012
ACTBlong
Q2	1.00 (Ref)
Q1	1.38	0.79–2.40	0.254
Q3	1.01	0.53–1.95	0.957

Abbreviations: HR= Hazard Ratio, CI=Confidence Interval, PR=Progesteron Receptor, MBC= Metastatic Breast Cancer, ACTB=Actin fragments, Q= Interquartile Range.

Fig. 2.

Kaplan-Meier plots for the impact on PFS of different ACTB_short Qs (A), ACTB_medium Qs (B) and ACTB_long Qs (C). Relative to the intermediate quartile group (Q2), a significant prognostic impact was noted for ACTB_short Q3 (HR 1.92, P = 0.041) and ACTB_medium Q3 (HR 0.47, P = 0.043). Abbreviations: T0, baseline; T3, after 3 months; T6, after 6 months; Q, Interquartile range.

Table 5.

PFS at 12, 24 and 36 months for the respective interquartiles of ACTB_short, ACTB_medium and ACTB_long and the corresponding 95 % confidence intervals. Abbreviations: CI= 95 % Confidence intervals; Q= Interquartile Range; ACTB= Actin fragments.

Progression free survival	12 months	24 months	36 months
ACTBshort
Q1	71 % (CI 52 %–84 %)	54 % (CI 35 %–70 %)	37.5 % (CI 21 %–54 %)
Q2	79.5 % (CI 67 %–88 %)	64.8 % (CI 56 %–71 %)	53.8 % (CI 39 %–67 %)
Q3	61.4 % (CI 42 %–76 %)	40 % (CI 19 %–60 %)	40 % (CI 19 %–60 %)
ACTBmedium
Q1	67.7 % (CI 55 %–78 %)	54.1 % (CI 35 %–69 %)	36.6 % (CI 18 %–55 %)
Q2	67.7 % (CI 55 %–78 %)	50.5 % (CI 37 %–62 %)	41.3 % (CI 28 %–54 %)
Q3	81.4 % (CI 63 %–91 %)	66.8 % (CI 43 %–82 %)	66.8 % (CI 43 %–82 %)
ACTBlong
Q1	64.5 % (CI 45 %–79 %)	54.7 % (CI 36 %–70 %)	35.4 % (CI 17 %–55 %)
Q2	75.9 % (CI 63 %–85 %)	56.5 % (CI 43 %–68 %)	47.5 % (CI 34 %–60 %)
Q3	72.4 % (CI 54 %–85 %)	51.6 % (CI 28 %–71 %)	51.6 % (CI 28 %–71 %)

Univariable analysis at baseline in terms of OS showed a significant impact on prognosis for patients with PR positive status (HR 0.28, CI 0.13–0.58, P < 0.001), lobular histotype (HR 2.99, CI 1.35–6.65, P = 0.007), liver metastases (HR 2.21, CI 1.07–4.56, P = 0.031), node metastases (HR 2.27, CI 1.04–4.97, P = 0.040) and ACTB_short Q3 (HR 3.94, 95 % CI 1.60–9.68 P = 0.003) (Fig. 3,Table 4). In multivariable analyses, PR positive tumors were associated with lower risk of death (HR 0.22, CI 0.09–0.52, P=0.001), while lobular histotype (HR 3.54, CI 1.44–8.73, P = 0.006), presence of node metastases (HR 3.25, CI 1.27–8.32, P=0.014) and ACTB_short Q3 (HR 3.25, 95 % CI 1.27– 8.32, P = 0.023) were associated with worse prognosis (Table 4). OS at 12, 24 and 36 months for the respective interquartiles of ACTB_short, ACTB_medium and ACTB_long are reported in Table 6.

Fig. 3.

Kaplan-Meier plots for the impact on OS of different ACTB_short Qs (A), ACTB_medium Qs (B) and ACTB_long Qs (C). Relative to the intermediate quartile group (Q2), a significant prognostic impact was noted for ACTB_short Q3 (HR 3.94, 95 % CI 1.60–9.68 P = 0.003) when compared to the intermediate quartile group (Q2). Abbreviations: T0, baseline; T3, after 3 months; T6, after 6 months; Q, Interquartile range.

Table 4.

Univariate and Multivariate Cox Regression Analysis for Overall Survival.

	Univariate analysis				Multivariate Analysis
	HR	95 % CI	p-value	HR		95 % CI	p-value
PR status
Negative	1.00 (Ref)			1.00 (Ref)
Positive	0.28	0.13–0.58	<0.001	0.22		0.09–0.52	0.001
De novo MBC
Yes	1.00 (Ref)
No	0.58	0.23– 1.48	0.258
Histotype
Ductal	1.00 (Ref)			1.00 (Ref)
Lobular	2.99	1.35–6.65	0.007	3.54		1.44–8.73	0.006
Liver Metastasis
No	1.00 (Ref)			1.00 (Ref)
Yes	2.21	1.07–4.56	0.031	1.97		0.90–4.33	0.091
Lung Metastasis
No	1.00 (Ref)
Yes	0.52	0.18–1.47	0.213
CNS Metastasis
No
Yes	not evaluable
Bone Metastasis
No	1.00 (Ref)
Yes	1.44	0.62–3.38	0.393
Node Metastasis
No	1.00 (Ref)			1.00 (Ref)
Yes	2.27	1.04–4.97	0.040	3.25		1.27–8.32	0.014
ACTBshort
Q2	1.00 (Ref)
Q1	1.49	0.60–3.68	0.391	1.64		0.61–4.37	0.326
Q3	3.94	1.60–9.68	0.003	3.74		1.20–11.67	0.023
ACTBmedium
Q2	1.00 (Ref)
Q1	1.03	0.44–2.42	0.939	1.81		0.72–4.54	0.207
Q3	0.93	0.33–2.57	0.888	0.31		0.08–1.24	0.098
ACTBlong
Q2	1.00 (Ref)
Q1	0.76	0.29–1.96	0.567
Q3	1.77	0.73–4.28	0.207

Abbreviations: HR= Hazard Ratio, CI=Confidence Interval, PR=Progesteron receptor, MBC= Metastatic Breast Cancer, ACTB=Actin fragments, Q= Interquartile Range.

Table 6.

OS at 12, 24 and 36 months for the respective interquartiles of ACTB_short, ACTB_medium and ACTB_long and the corresponding 95 % confidence intervals. Abbreviations: CI= 95 % Confidence intervals; Q= Interquartile Range; ACTB=Actin fragments.

Overall Survival	12 months	24 months	36 months
ACTBshort
Q1	94 % (CI 77 %–98 %)	81 % (CI 63 %–91 %)	72 % (CI 51 %–85 %)
Q2	100 %	91.2 % (CI 80 %–96 %)	78.8 % (CI 63 %–88 %)
Q3	87 % (CI 69 %–95 %)	64 % (CI 37 %–81 %)	44 % (CI 17 %–68 %)
ACTBmedium
Q1	90.3 % (CI 73 %–97 %)	77.2 % (CI 58 %–88 %)	72.0 % (CI 51 %–85 %)
Q2	96.7 % (CI 88 %–99 %)	86.3 % (CI 74 %–93 %)	70.7 % (CI 56 %–82 %)
Q3	96.7 % (CI 79 %–99 %)	78.6 % (CI 50 %–92 %)	67.4 % (CI 34 %–86 %)
ACTBlong
Q1	93.6 % (CI 77 %–98 %)	87.1 % (CI 69 %–95 %)	76.9 % (CI 54 %–89 %)
Q2	96.8 % (CI 88 %–99 %)	82.7 % (CI 70 %–90 %)	73.5 % (CI 59 %–84 %)
Q3	93.8 % (CI 77 %–98 %)	74.9 % (CI 47 %–90 %)	47.6 % (CI 19 %–72 %)

Discussion

In the context of luminal MBC, liquid biopsy features such as ctDNA, CTCs, epigenetics and exosomes are gaining traction as tools to investigate and monitor tumor evolution [10]. In this setting, fragmentomics is emerging as a new parameter capable of augmenting the insights obtainable through ctDNA [6].

At baseline, the distribution of ACTB_short changed significantly with respect to PR status and the presence of liver metastases. The highest interquartile group of ACTB_short was inversely associated with PR expression and directly associated with the presence of liver metastases. Given that both PR-negative status and the presence of liver metastases are acknowledged as unfavorable prognostic indicators in luminal-like MBC, these findings suggest these conditions are concurrently characterized by high tumor shedding, adding a further biological rationale to their aggressiveness.

Consistent with our previous work, where a >20 % increase between baseline and first evaluation of ACTB_short demonstrated a significant negative prognostic impact in PFS [7], the highest interquartile of baseline ACTB_short showed an unfavorable impact on both PFS and OS, confirming the association between increased tumor shedding and heightened aggressiveness, suggesting that patients at risk of faster progression could be identified not only through dynamic but also with baseline assessment.

Although not statistically significant, an intriguing trend towards an unfavorable prognostic effect was observed for the lowest interquartile of ACTB_short, hinting at the existence of a subgroup of non-shedding yet equally aggressive tumors. Considering that this subgroup was not significantly associated with the presence of brain metastases, which, although known as non-shedding sites, are associated with poorer outcomes, it is noteworthy that these patients, characterized by low shedding profile, require further clinical and biological analyses to be fully characterized. Future analyses through high sensitivity NGS methods may help identify potential concomitant mutations or associated epigenetic patterns, contributing to define a comprehensive biological landscape. Notably, for such cases, the detection of ctDNA during subsequent timepoints could serve as a crucial indicator of disease progression but further research will be necessary to delve into this hypothesis [11]

Interestingly, a favorable impact on PFS, but not on OS, was observed for the highest interquartile of ACTB_medium. This intermediate-length fragment, represented by ACTB_medium, is associated with neutrophils shedding. The impact on PFS alone may suggest a treatment-specific relationship, aligning with literature indicating an increased proportion of white blood cells during treatment with CDK4/6 inhibitors as a significant favorable prognostic factor [12].

To date, plasma ctDNA is gaining momentum as an emerging surrogate for tumor burden not only due to its prognostic value [[13], [14], [15]] but also for its predictive role [16,17]. However, most proposed approaches rely on either single-gene or gene panels, thereby restricting their monitoring sensitivity to the number of analyzed genes. Here, we demonstrate the feasibility of a droplet digital PCR-based agnostic approach focused on fragments size, which maximizes transferability and cost-effectiveness. While other fragmentomics approaches, such as nucleosomal patterns [18], end-fragment signatures [19], and epigenomic modifications [18,20,21], have been proposed as alternative methods, the evaluation of ACTB fragments currently emerges as one of the most scalable, cost-effective and robust strategies in the context of MBC.

Therefore, while traditional methods based on ddPCR and NGS technologies remain crucial for deciphering tumor clonal evolution and identifying potential target genes and resistance mechanisms, the assessment of ACTB fragments may serve as a complementary approach. This method, independent of the tumor’s mutational pattern, allows for both short- and long-term follow-up regardless of clonal and subclonal tumor evolution, potentially enabling the early identification of patient subgroups at risk of early progression. Alongside other features analyzable through ctDNA, analysis of progression hazard distribution peaks could facilitate the personalization of radiological monitoring modalities and schedules, adapting them to the varying baseline risk [[22], [23], [24]].

This study, however, has some limitations. Although it was a prospective study that enrolled patients receiving first-line treatment for luminal MBC, the relatively small sample size and the homogeneity of the study population necessitate validation of this approach on a larger scale with better representation of different subgroups, each potentially exhibiting specific shedding characteristics. Moreover, for some patients, follow-up is still in its early stages, potentially underestimating the prognostic impact of the biomarkers considered. The analysis of subsequent time points, particularly those preceding and concurrent with progression, will enable a better understanding of the biological behavior and dynamics of circulating actin fragments.

Conclusion

This study demonstrates the feasibility of employing a fragmentomics approach in cell free DNA based on the ACTB gene, in the context of first-line treatment of luminal-MBC patients. Although mutation-agnostic, this quantitative approach shows promise to being integrated into clinical practice, aiming to understand and predict the tumor dynamics.

Ethics approval and consent to participate

The study was approved by the ethics committee under the CEUR-2018-Sper-056-CRO protocol. The patients/participants provided their written informed consent to participate in this study. Written informed consent was obtained from the individual(s) for the publication of any potentially identifiable images or data included in this article. The study was performed in concordance with the Health Insurance Portability and Accountability Act and the Declaration of Helsinki.

Authors' contributions

LF, AF, GD, LG, FP conceived and supervised the study; LF, AF, EM, AD, DB, LC, SB, CN, SR,ER, SB, BP,SD, LA, MB, AM, BB, GD, LG, FP enrolled patients and collected follow-up patient information; AF, EM, BB, GD and LG performed molecular analysis; LG, AF, FG and LG performed the bioinformatics analysis; LG, AF, FG and LG performed the statistical analysis; EM, AD, DB, LC, SB, CN, SR,ER, SB, BP,SD, LA, MB, AM, BB, GD, LG, FP reviewed the paper and provided critical data interpretation; LF and AF wrote the manuscript with input from all authors. All of authors have read and approved the manuscript.

CRediT authorship contribution statement

Lorenzo Foffano: Writing – review & editing, Resources, Formal analysis, Visualization, Investigation, Conceptualization, Writing – original draft, Methodology, Data curation. Alessandra Franzoni: Visualization, Methodology, Formal analysis, Writing – original draft, Resources, Funding acquisition, Conceptualization, Writing – review & editing, Validation, Investigation, Data curation. Elisabetta Molteni: Writing – review & editing, Methodology, Validation, Writing – original draft. Fabiola Giudici: Writing – original draft, Investigation, Writing – review & editing, Methodology, Data curation, Software, Formal analysis. Arianna Dri: Writing – review & editing, Resources. Debora Basile: Writing – review & editing, Investigation, Resources. Linda Cucciniello: Resources, Writing – review & editing. Silvia Buriolla: Resources, Writing – review & editing. Claudia Noto: Resources, Writing – review & editing. Stefania Russo: Resources, Writing – review & editing, Data curation. Elena Nascimbeni: Resources, Writing – review & editing, Project administration. Silvia Bolzonello: Writing – review & editing, Resources. Brenno Pastò: Resources, Writing – review & editing. Serena Della Rossa: Resources, Writing – review & editing. Lorenzo Allegri: Resources, Writing – review & editing, Data curation. Marta Bonotto: Resources, Writing – review & editing, Methodology. Alessandro Marco Minisini: Writing – review & editing, Resources. Barbara Belletti: Writing – review & editing, Data curation, Resources. Giuseppe Damante: Writing – review & editing, Methodology, Resources. Lorenzo Gerratana: Writing – review & editing, Methodology, Formal analysis, Resources, Funding acquisition, Conceptualization, Writing – original draft, Investigation, Data curation. Fabio Puglisi: Writing – review & editing, Funding acquisition, Methodology, Conceptualization, Resources, Formal analysis.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:

M.B. reports advisory/consultancy fee from AstraZeneca, Lilly, MSD, Novartis, Pfizer, SeaGen; travel grants from Lilly, Roche. A.M. reports advisory/consultancy fee from Novartis, MSD, BMS, Merk, Sunpharma, PierreFabre, Gilead, Seagen, Genomic Health; invited speech from Novartis, MSD, BMS, Merk, Sunpharma, Sanophi, PierreFabre, AstraZeneca, Daiichi Sankyo; travel grants from Gilead, PierreFabre; other familial: MSD, AstraZeneca, Pharmamar, GSK. L.G. reports advisory/consultancy fee from AstraZeneca, Daiichi Sankyo, Eli Lilly, GlaxoSmithKline, Incyte, Novartis, Pfizer, Merck Sharp & Dohme, Menarini Stemline, Abbvie; research funding from Menarini Silicon Biosystems. F.P. reports honoraria for advisory boards, activities as a speaker, travel grants, research grants from Amgen, Astrazeneca, Daiichi Sankyo, Celgene, Eisai, Eli Lilly, Exact Sciences, Gilead, Ipsen, Italfarmaco, Menarini, MSD, Novartis,Pierre Fabre, Pfizer, Roche, Seagen, Takeda, Viatris; Research funding from Astrazeneca – Eisai – Roche

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.