High mid-treatment tumour RNA disruption in patients with HER2-negative breast cancer is associated with improved disease-free survival after neoadjuvant chemotherapy

Abstract

Background

High tumour ribosomal RNA degradation (RNA disruption) during neoadjuvant chemotherapy has been associated with a post-treatment pathologic complete response (pCR) and improved disease-free survival (DFS) in breast cancer patients. We further assessed the relationship between tumour RNA disruption or other metrics and neoadjuvant chemotherapy outcome using data from the NeoAva clinical trial (NCT00773695).

Methods

Patients with early HER2-negative breast cancer received FEC-T chemotherapy ± bevacizumab in a randomized fashion. Biopsies were taken pre-treatment and after 12 and 25 weeks of chemotherapy. RNA and proteins extracted from the biopsies were used to compute the RNA disruption index (RDI) and to quantify levels of 210 proteins using protein array analysis at 12 weeks.

Results

Tumour RDI values were higher mid- and post-treatment than pre-treatment (p < 0.0001). Patients with tumour RDI values > 1.1 exhibited higher disease-free and breast cancer-specific survival than patients with RDI values ≤ 1.1 (p = 0.049 and 0.031, respectively). While RDI values were higher for patients on the bevacizumab-containing regimen (p = 0.003), this was not associated with improved survival. Survival on either regimen was not significantly associated with a post-treatment pCR or an improved residual cancer burden (RCB) score. Significant differences in apoptotic, EMT, Notch, G1-S checkpoint, and DNA damage response pathways were seen between high- and low-RDI tumours.

Conclusions

High tumour RNA disruption during neoadjuvant chemotherapy was associated with improved DFS and may better predict outcome than the post-treatment pCR rate or RCB. If validated as an independent predictor of chemotherapy outcome, RNA disruption assessments during treatment may prove informative in making treatment escalation or de-escalation decisions.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13058-025-02092-9.

Background

We have observed that various mechanistically distinct chemotherapy drugs can induce ribosomal RNA (rRNA) degradation in tumour cells in vitro and in vivo [1–7]. This phenomenon, termed “RNA disruption”, is reproducibly associated with tumour cell death [8, 9]. RNA disruption can be quantified in biological samples using the RNA disruption assay (RDA), which involves isolating total RNA, resolving the various RNAs by capillary gel electrophoresis, and quantifying the extent of RNA disruption by dividing the combined areas of all abnormal peaks on RNA electropherograms by the combined areas of the 28 S and 18 S rRNAs (the RNA disruption index or RDI).

In the NCIC-CTG-MA.22 clinical trial [10], high mid-treatment tumour RDI values in patients with breast cancer were associated with a high post-treatment pathologic complete response (pCR) rate and improved disease-free survival (DFS) [3]. High RNA disruption was observed in all tumour subtypes [defined by their expression of the human epidermal growth factor 2 (HER2), estrogen receptor (ER), and progesterone receptor (PR)].

There is an urgent unmet need to reduce breast cancer overtreatment, since the benefit of neoadjuvant chemotherapy varies not only according to subgroups, but also among individual patients. In addition to toxic treatment side-effects [11], such patients may experience early tumour progression and poor outcome [12, 13]. New tools are needed to identify, during chemotherapy, patients at high risk of treatment failure, who could possibly forgo further cycles of the ineffective regimen and move rapidly to alternate treatments. The RDA may be particularly useful for this approach (termed “response-guided neoadjuvant chemotherapy”) [14]. Consistent with this potential, high tumour RNA disruption after one cycle of neoadjuvant chemotherapy was associated with a higher pCR rate in patients with HER2+/ER- breast cancer [6].

The NeoAva clinical trial (NCT00773695) investigated the efficacy of adding bevacizumab to neoadjuvant chemotherapy in 132 eligible patients with early primary HER2- breast cancer. Patients received 5-fluorouracil, epirubicin and cyclophosphamide (FEC) for 12 weeks followed by taxane (T) therapy (paclitaxel or docetaxel) for 12 weeks. Half of the patients (66 patients) also received concurrent bevacizumab for 24 weeks. Co-administration of bevacizumab increased the pCR rate from 5 to 20% of patients with ER + tumours [15]. pCR was defined as the absence of tumour cells in both the surgical specimen and axillary lymph nodes [residual cancer burden (RCB) score 0]. The addition of bevacizumab also improved disease-free survival rates among good responders (RCB 0 and 1) [16]. Tumour transcriptional profiling suggested that the improved treatment response was associated with elevated immune activity and the expression of specific genes [15, 16]. Changes in gene copy number [17], the expression of specific miRNAs [18], serum cytokine levels [19], and a nine-protein signature [20] were all found to be associated with response to the bevacizumab-containing regimen.

This report utilizes RCB data, clinical response data, and RNA capillary electrophoresis data to compare the ability of RCB, the post-treatment pCR rate and on-treatment RDI values to predict patient outcome [DFS and breast cancer-specific survival (BCSS)] after neoadjuvant FEC-T chemotherapy ± bevacizumab. It also examines whether the expression of proteins involved in key oncogenic signaling pathways is significantly different between high-RNA disruption and low-RNA disruption tumours.

Methods

Study design

Women with non-metastatic operable HER2- breast cancer with tumours ≥ 2.5 cm (n = 132) were recruited from Oslo University Hospital (Oslo, Norway) and St. Olav’s Hospital (Trondheim, Norway) between November 2008 and July 2012. All patients received neoadjuvant treatment with four cycles of 5-fluorouracil 600 mg/m², epirubicin 100 mg/m² and cyclophosphamide 600 mg/m² every three weeks (FEC), followed by either docetaxel 100 mg/m² every three weeks for 12 weeks, or weekly infusions of paclitaxel 80 mg/m² for 12 weeks (T). Half of the patients were randomized to receive bevacizumab (15 mg/kg every three weeks in patients receiving FEC + docetaxel, or 10 mg/kg every two weeks in patients receiving FEC + paclitaxel). Tumour biopsies were taken from patients pre-treatment (screening), after four cycles of FEC chemotherapy (at 12 weeks), and after the final taxane dose (at 25 weeks). Biopsies were flash-frozen in liquid nitrogen and stored at -80 °C. The biopsies, tumour receptor expression data, and patient clinical response data were then used to identify putative metrics or biomarkers for prediction of patient outcome after neoadjuvant FEC-T chemotherapy (± bevacizumab). Treatment response was assessed by computing pCR rate or the RCB score at surgery using the MD Anderson RCB calculator. Disease-free and overall survival were measured, with a median follow-up time of 6.7 years. The study was approved by the institutional protocol review board, the regional ethics committee, the Norwegian Medicines Agency and carried out in accordance with the Declaration of Helsinki, International Conference on Harmony/Good Clinical practice. The study is registered in the http://www.ClinicalTrials.gov/ database with the identifier NCT00773695.

RNA isolation

Total RNA was isolated from flash-frozen tumour biopsies using the AllPrep DNA/RNA Mini kit (Qiagen) run on the QIAcube automated platform (Qiagen). The RNAs within the samples were then resolved by capillary gel electrophoresis using the 2100 Bioanalyzer (Agilent Technologies) with the RNA 6000 Nano kit (Agilent Technologies).

RNA disruption assay

RDA analysis of the RNA samples was performed using electropherogram data stored in xad files on the 2100 Bioanalyzer. Using a proprietary algorithm developed by Rna Diagnostics, RDI values were computed for each sample with detectable levels of both the 28S and 18S rRNAs. RNA concentrations were computed for each sample by comparing the total area under the curve of the sample’s electropherogram to that of the RNA 6000 Ladder (Agilent Technologies) included in each run.

Reverse-phase protein array (RPPA) pathway analysis

The reverse-phase protein array (RPPA) core facility at the MD Anderson Cancer Center (Houston, TX) quantified the expression of 210 proteins in lysates from freshly frozen tumour biopsies collected at week 12 of treatment. In this exploratory study, only tumours with both an evaluable RPPA protein profile and an RDI score were considered for further analysis (n = 77). RPPA pathway Z score analysis was used to measure the activity of key oncogenic signaling pathways in the tumours. All positively associated predictors were summed minus the predictors that were negatively associated with the pathway (Supplemental Table 1).

Statistical analysis of data

All datasets (except the RPPA dataset) were not normally distributed and were both positively skewed and leptokurtic. There was unequal variance between compared groups, indicating that the data was heteroscedastic. Thus, non-parametric two-tailed Mann-Whitney tests were conducted. Kaplan-Meier curves were used to assess differences in survival amongst groups. Univariate hazard ratios (HRs) and confidence intervals (CIs) were also computed. For RPPA analysis, t-tests were performed to assess the significance of differences between the high- and low-RDI groups. The small sample size of the NeoAva clinical trial precluded the use of multivariate cox regressions to identify which variables and proteins are strongly associated with patient survival. Statistically significant differences in various parameters between groups of patients are indicated in the text or depicted in figures as follows: (*, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001).

Results

Patient and tumour characteristics, and response to neoadjuvant chemotherapy

As shown in the consolidated standards of reporting trials (CONSORT) diagram (Fig. 1), 71 and 67 patients were randomly assigned to receive neoadjuvant FEC-T chemotherapy with and without bevacizumab, respectively. After randomization, three patients did not meet all eligibility criteria, and three discontinued treatment. Thus, in the primary endpoint analysis, 66 patients received FEC-T without bevacizumab, and 66 patients received FEC-T with bevacizumab. In our RDA analysis, additional patients (and their associated data) were excluded due to missing data, missing samples, or insufficient tumour RNA for RDI computation. This left 109, 98, and 106 patients for RDA analysis at screening, and at 12 and 25 weeks of treatment, respectively. Eighty-five per cent, 88%, and 87% of patients had ER + tumours at screening, and at 12 and 25 weeks of treatment, respectively.

Fig. 1.

Consolidated standards of reporting trials (CONSORT) diagram for the NeoAva clinical trial. 138 patients with HER2- tumours were treated with neoadjuvant FEC-T chemotherapy and an aromatase inhibitor, when tumours were ER+. Seventy-one patients were randomly assigned to also receive bevacizumab (Bev.) with FEC-T treatment, while 67 patients were only treated with FEC-T. Of these patients, 132 were assessed for clinical response to the regimens, while 109, 98, and 106 patients contributed biopsies for biomarkers studies prior to treatment (screening), after 12 weeks of treatment, and after 25 weeks of treatment, respectively. At the various timepoints, various metrics or biomarkers (including mid-treatment tumour RNA disruption, post-treatment pCR rate or post-treatment RCB) were assessed for their ability to predict FEC-T treatment outcome (DFS and BCSS) with and without additional bevacizumab treatment

Consistent with the full NeoAva dataset [15], the addition of bevacizumab to the FEC-T regimen increased the pCR rate (Table 1). Twelve of the 48 bevacizumab-treated patients (25%) assessed by RDA at 12 weeks achieved a pCR, compared to six of 50 for FEC-T only patients (12%). This finding did not reach statistical significance (p = 0.121, Fisher’s Exact test), likely due to the limited number of patients achieving a pCR.

Table 1.

NeoAva patients with RDA results by tumour ER expression status, bevacizumab treatment status, and by treatment response outcomes

Number of Patients	Screening	12 weeks	25 weeks
Total	109	98	106
ER+	93	86	92
ER-	16	12	14
No Bevacizumab	57	50	55
Bevacizumab	52	48	51
RCB Class 0	19	18	14
RCB Class 1	11	10	13
RCB Class 2	62	53	59
RCB Class 3	17	17	20
pCR	19	18	14
pCR No Bevacizumab	7	6	5
pCR Bevacizumab	12	12	9
No pCR	90	80	92

Tumour RNA disruption increased upon treatment and was associated with reduced RNA concentration

The RDA was performed on biopsies collected at screening (n = 109), and after 12 weeks (n = 98) and 25 weeks (n = 106) of treatment. Mean RDI values were 1.7 ± 0.2 at screening, 2.8 ± 0.3 after 12 weeks of treatment, and 3.2 ± 0.4 after 25 weeks of treatment (Fig. 2A). Thus, tumour RDI values during or upon completion of chemotherapy were significantly higher than those measured prior to chemotherapy (p < 0.0001 for both the 12- and 25-week biopsies; Mann-Whitney test). Average tumour RNA concentrations (Fig. 2B) were lower during (12 weeks) or after (25 weeks) chemotherapy (69.7 ± 11.5 ng/µL and 55.8 ± 9.2 ng/µL, respectively) than at screening (217.5 ± 22.7 ng/µL) (p < 0.0001 for both the12- and 25-week biopsies; Mann-Whitney test).

Fig. 2.

Mean RDI value and RNA concentration for patient biopsies at specific timepoints. Tumour biopsies were collected from all patients prior to treatment initiation (screening), and after 12 and 25 weeks of treatment. The extent of tumour RNA disruption was assessed using the RDA (panel A), and the tumour RNA concentration was measured using the 2100 Bioanalyzer (panel B). Data are presented as means ± S.E.M. Tumour RDI values at 12 weeks and at 25 weeks were significantly higher than those measured prior to chemotherapy (p < 0.0001 for both the 12- and 25-week biopsies; Mann-Whitney test). Average tumour RNA concentrations were significantly lower during and after chemotherapy than at screening (p < 0.0001 for 12- and 25-week biopsies; Mann-Whitney test)

Thus, the FEC regimen, with or without taxanes or bevacizumab, augments tumour RNA disruption and reduces tumour RNA content. There were higher proportions of patients with non-assessable (n/a) RNA samples (no RDA results) when biopsies were taken during or after chemotherapy (8.4% and 10.2% of patients at 12 and 25 weeks of treatment, respectively) than at screening (4.4% of patients) (Table 2).

Table 2.

Number (percentage), pCR status, and RCB status of total patients with not-assessable (n/a) RNA samples at each biopsy timepoint

Timepoint	Number of Patients with n/a Samples (% of total patients)	pCR Status of n/a Patients	RCB Status of n/a Patients
Screening	5 of 114 (4.4%)	pCR – 1 patient no pCR – 4 patients	RCB0 –1 patient RCB1 –1 patient RCB2 –2 patients RCB3 –1 patient
12 weeks	9 of 107 (8.4%)	pCR – 0 patients no pCR – 9 patients	RCB0 –0 patients RCB1 –1 patient RCB2 –7 patients RCB3 –1 patient
25 weeks	12 of 118 (10.2%)	pCR – 2 patients no pCR – 10 patients	RCB0 –2 patients RCB1 –1 patient RCB2 –9 patients RCB3 –0 patients

Tumour RNA disruption after 12 weeks of treatment correlates with clinical outcome

Patients with a RCB class of 0 or 1 post-treatment had a 2-fold higher median (M) tumour RDI score at 12 weeks (n = 28; M = 2.6) than patients with a RCB class of 3 (n = 17; M = 1.3) (p = 0.006; Mann-Whitney test) (Table 3; Fig. 3).

Table 3.

Mean and median RDI values by RCB class

	RCB Class 0 (pCR)	RCB Class 1	RCB Classes 0/1	RCB Class 2	RCB Class 3	No pCR
Mean RDI Value (± SEM)	3.5 ± 0.7	3.4 ± 0.7	3.5 ± 0.5	2.8 ± 0.4	1.8 ± 0.3	2.7 ± 0.3
Median RDI Value (IQR)	2.6 (1.6–4.2)	2.9 (1.4–6.1)	2.6 (1.5–4.5)	2.0 (1.1–3.2)	1.3 (1.0-2.4)	1.8 (1.1–3.3)
Number of Patients	18	10	28	53	17	80

Fig. 3.

Maximum tumour RNA disruption measured in 98 patients at 12 weeks and its association with clinical outcome as measured by RCB class. The maximum tumour RDI value at 12 weeks for each patient that had either a post-treatment RCB score in class 3 or a post-treatment RCB score in classes 0 or 1 was plotted against the corresponding tumour RNA concentration at 12 weeks. A 12-week tumour RDI value of 1.1 (dotted line) was then used to differentiate between these two groups of patients

pCR recipients had a median RDI value of 2.6 (n = 18), which was higher than the median RDI value of 1.8 for patients with residual disease (n = 80); this finding approached, but did not reach, statistical significance (p = 0.090, Mann-Whitney test). Patients with ER- tumours had higher levels of RNA disruption at 12 weeks (n = 12, M = 3.0) than patients with ER + tumours (n = 86, M = 1.8; Table 4), though this finding did not reach statistical significance. Patients with ER- tumours also had a higher pCR rate (6 of 12, 50%) than patients with ER + tumours (12 of 86, 14%), consistent with previously published findings [13].

Table 4.

Mean and median RDI values by ER status

	Mean RDI (± SEM)	Median RDI	pCR Rate
ER+	2.7 ± 0.3	1.8* (IQR 1.1–3.3)	12 of 86 (14%)
ER-	3.7 ± 0.7	3.0* (IQR1.9-5.4)	6 of 12 (50%)

The distribution of patients by RCB class or pCR using an RDI cut point value of 1.1 is shown in Table 5. The cut point of 1.1 was employed since ≥93% of patients with mid-therapy RDI values > 1.1 achieved a pCR or were in RCB classes 0 or 1 post-treatment.

Table 5.

Distribution of patients within RCB classes and by pCR using an RDI cutoff value of 1.1

	RCB 0/1 (n, %)	RCB 2 (n, %)	RCB 3 (n, %)	pCR (n, %)	No pCR (n, %)	All Patients (n, %)
RDI ≤ 1.1	2 (7%)	15 (28%)	6 (35%)	1 (6%)	22 (28%)	23 (23%)
RDI > 1.1	26 (93%)	38 (72%)	11 (65%)	17 (94%)	58 (72%)	75 (77%)
Total	28	53	17	18	80	98

Comparative Kaplan-Meier survival curves for patients with tumour RDI values > 1.1 or ≤ 1.1 below are also depicted in Fig. 4. Patients with a 12-week RDI value ≤ 1.1 were highly unlikely to achieve a pCR (negative predictive value of 96%). RDI values of biopsies collected at screening or after 25 weeks of treatment could not be used to predict RCB class or pCR. The DFS of patients with a 12-week tumour RDI > 1.1 was significantly greater than that of patients with an RDI ≤ 1.1 at 12 weeks (p = 0.049; hazard ratio (HR) = 2.6; 95% confidence interval (CI) = 0.8–8.1; Fig. 4A). Similarly, BCSS was higher for patients with a 12-week tumour RDI > 1.1 compared to patients with a tumour RDI ≤ 1.1 at 12 weeks (p = 0.031; HR = 3.1; 95% CI 0.8–11.7; Fig. 4B). No difference in DFS was noted between patients that achieved a post-treatment pCR and patients with residual disease (for the RDA-assessable 12-week subgroup). These patients had predominantly ER+/HER2- tumours, which are known to have reduced pCR rates [13]. No association between increased pCR rates in the presence of bevacizumab and improved DFS was observed, consistent with findings for the full NeoAva cohort [15]. Higher survival could not be predicted based on tumour RDI values at screening (Fig. 4C) or after 25 weeks of chemotherapy (post-treatment; Fig. 4D).

Fig. 4.

Disease-free and breast cancer-specific survival curves for NeoAva patients using various clinical response metrics. (A) Patients with a 12-week tumour RDI value > 1.1 had statistically significant improved DFS relative to patients with a 12-week tumour RDI value ≤ 1.1 (p = 0.049) (HR = 2.6; 95% CI 0.8–8.1). (B) Patients with a 12-week tumour RDI value > 1.1 had significantly improved BCSS relative to patients with a 12 week tumour RDI value ≤ 1.1 (p = 0.031) (HR = 3.1; 95% CI 0.8–11.7). (C) Patients with a pre-treatment tumour RDA value > 1.1 had no significant difference in DFS relative to patients with a 12-week tumour RDI value ≤ 1.1. (D) Patients that achieved a pCR post-treatment did not demonstrate improved DFS relative to patients without a pCR (having residual disease)

Biological processes associated with low RNA disruption

Preliminary exploratory RPPA studies utilizing available 12 week biopsies of sufficient size from 77 of the 132 patients were conducted to identify differences in the expression of proteins between high-RNA disruption (RDI > 1.1) and low-RNA disruption (RDI ≤ 1.1) tumours. As shown in Figs. 5 and 12-week tumours with an RDI < 1.1 had a significantly lower apoptotic balance (p = 0.001), epithelial-mesenchymal transition (EMT) (p < 0.001), and notch signaling (p < 0.001). These tumours also had increased expression of G1-S checkpoint-related proteins (p = 0.001) and DNA damage response (DDR) proteins (p = 0.001). Other pathways were not significantly associated with the RDI score [Supplemental Data Files 1 (plots) and 2 (data)].

Fig. 5.

RPPA pathway scores of apoptotic balance, epithelial-mesenchymal transition (EMT), notch signaling, G1-S checkpoint, and DNA damage response in tumours with high (grey, n = 57) and low (red, n = 20) RDI values at 12 weeks of treatment

Bevacizumab treatment resulted in increased RNA disruption measured at 12 weeks

The median tumour RDI value for patients undergoing FEC-T chemotherapy with bevacizumab was significantly higher than that obtained for patients given FEC-T alone (Fig. 6, p = 0.003, Mann Whitney test). No significant differences in RDI values were noted between these patients when samples were taken pre-treatment or at 25 weeks. Bevacizumab treatment increased the pCR rate both in RDA-assessable patients at 12 weeks (12 pCRs in 48 FEC-T + bevacizumab patients, versus 6 pCRs in 50 FEC-T patients; Table 1) and in the full NeoAva dataset [15]. Also, like prior findings [16], bevacizumab treatment resulted in fewer patients in RCB 3 (6 of 48 patients) compared to FEC-T only patients (11 of 50 patients). Nevertheless, these changes in the pCR rate or RCB score did not result in improved DFS [16].

Fig. 6.

Increased tumour RDI values upon addition of bevacizumab to the FEC-T regimen for NeoAva patients after 12 weeks. RDI values in patients treated with FEC-T and bevacizumab (FEC-T + Bev.) were significantly higher than in patients that received treatment with FEC-T only (median value of 2.4 compared to 1.4; Mann-Whitney Test, p = 0.003)

Tumour RNA Disruption at 12 weeks and treatment response for patients With ER+/Her- disease

Since the vast majority of patients in the NeoAva trial had ER+/HER2- tumours, we then restricted our analysis to patients with ER+/HER2- tumours only. In this analysis, the median tumour RDI value at 12 weeks was also significantly higher for patients with a RCB class of 0 or 1 (2.6) than for patients with a RCB class of 3 (1.3), (Table 3, p = 0.03 Mann Whitney test). As shown in Supplemental Fig. 1, DFS for patients with ER+/HER2- tumours was also higher for patients with tumour RDI values > 1.1 compared to patients with RDI values ≤ 1.1 (p = 0.02, HR 3.2, 95% CI 0.9 to 10.8). Similarly, BCSS for patients with ER+/HER2- tumours was higher for patients with tumour RDI values > 1.1 compared to patients with RDI values ≤ 1.1 (Supplemental Fig. 2; p = 0.008, HR 4.7, 95% CI 1.1 to 19.8).

Similar to the larger cohort of NeoAva patients with HER2- tumours, our analysis of patients with ER+/HER2- tumours revealed no significant differences in DFS between patients with RCB 3 and patients with RCB 0 or 1 or between patients with a post-treatment pCR and patients without a pCR. DFS was also not significantly different between bevacizumab-treated patients and non-treated patients.

Discussion

Relationship between tumour RNA disruption and treatment outcome

The high tumour RDI values at 12 weeks in patients with a post-treatment pCR suggests an association between high tumour RNA disruption and improved treatment outcome, although this finding was not statistically significant (Table 3). Nevertheless, the median tumour RDI value at 12 weeks for responders to treatment (patients in RCB classes 0 and 1, n = 28, M = 2.6) was significantly greater than that of non-responding patients in RCB class 3 (n = 17, M = 1.3; p = 0.006, Mann-Whitney test, Table 3). Moreover, although the dataset is small, there appeared to be an inverse association between the level of RCB and the mean tumour RDI value (Table 3). As a percentage, there were about 5-fold more patients that had tumour RDI values ≤ 1.1 in RCB class 3 than patients in RCB classes 0 or 1. There were also about twice the percentage of patients that had tumour RDI values > 1.1 in RCB classes 0 or 1 than patients in RCB class 3 (Table 5). Patients with tumour RDI values > 1.1 had significantly greater DFS than patients with tumour RDI values ≤ 1.1 (Fig. 4A and B).

The higher tumour RDI scores and the lower tumour RNA content at 12 and 25 weeks (Fig. 2A and B) were likely due to chemotherapy administration, since there was no difference in DFS between patients with pre-treatment tumour RDI values > 1.1 and patients with pre-treatment tumour RDI values ≤ 1.1 (Fig. 4C). Moreover, the ~ 2-fold change in mean tumour RDI values and ~ 4-fold change in mean tumour RNA concentrations between pre-treatment and on-treatment tumours (with significant and very low p values of < 0.0001) help to illustrate the strong effect that chemotherapy treatment has on the rRNAs. Nevertheless, while tumour RDI values were highest after 25 weeks of chemotherapy (Fig. 2A), they could not be used (like 12- week biopsies) to predict survival post-treatment. We have shown that tumour cellularity within biopsies decreases with increasing time of treatment [1]. The reduced tumour cellularity and infiltration of the tumour with normal tissue at the end of treatment may mute our ability to quantify chemotherapy-induced tumour RNA disruption. This may help to explain why the RDA was unable to predict treatment outcome using tumour biopsies from NeoAva patients after 25 weeks of treatment.

The RDA may be superior to post-treatment pCR and RCB assessments in predicting treatment outcome

Approximately half of breast cancer patients are currently treated with systemic chemotherapy involving cytotoxic and targeted chemotherapy agents [21–23] and according to data from Kaiser Permanente Northern California and Kaiser Permanente Washington, the use of neoadjuvant chemotherapy has increased in the United States from 4.1 to 14.7% of breast cancer patients from 2006 to 2019 [24]. Similarly, neoadjuvant chemotherapy usage in Europe for the treatment of clinically node-positive breast cancer has increased from 20 to 46.7% of patients between 2018 and 2022 [25]. While the use of chemotherapy for the treatment of ER + breast cancer has clearly decreased due to genomic testing, a study in 2018 revealed that, nevertheless, about 21.3% of patients with ER + disease undergo chemotherapy. These patients typically have high risk, node-positive disease [26]. There is a strong association between a pCR after neoadjuvant chemotherapy and increased DFS for patients with specific subtypes of breast cancer [13]. In a recent meta-analysis [27], there was no overlap in 95% CIs for DFS curves between pCR recipients (higher survival) and those with disease recurrence (lower survival) in patients with HER2-/HR- breast cancer. In contrast, some degree of overlap was observed between the pCR and disease recurrence groups in patients with HER2+/ER- breast cancer. The degree of overlap was greatest for patients with ER+/HER2- breast cancer, where no significant difference in DFS could be observed between the two groups [27]. Similarly, in the small NeoAva cohort of patients with predominantly ER+/HER2- tumours, no significant difference in DFS curves was observed between the pCR and non-PCR groups (Fig. 4D). There is increasing evidence that the post-treatment pCR rate is not an ideal predictor of outcome after neoadjuvant chemotherapy. Supporting this view, very recent studies using highly sensitive, tumour-informed ctDNA assays suggest that post-treatment detection of circulating tumour DNA (ctDNA) is superior to pCR rate in predicting disease recurrence in breast cancer patients, in particular for patients with triple negative disease [28, 29].

DFS curves were also not significantly different amongst NeoAva patients of various RCB classes, although survival was lowest for RCB class 3 patients (Supplemental Fig. 3). In contrast, patients with high RDI values at 12 weeks (RDI > 1.1) had significantly higher DFS and BCSS than patients with low RDI values (RDI ≤ 1.1; Fig. 4A and B). Thus, mid-treatment RDA tests may prove to be superior to post-treatment pCR and RCB assessments in predicting outcome after neoadjuvant chemotherapy.

The 12-week biopsies were taken before the administration of taxanes, suggesting that the amount of RNA disruption at 12 weeks of FEC chemotherapy was sufficient to demonstrate an association with DFS after FEC-T chemotherapy (± bevacizumab). It should be noted that the 1.1 cut point used in this study was empirically derived and was different than that used in the MA.22 clinical trial [3]. This likely reflects differences in RNA isolation methods which we have found has an impact on the measurement of RNA disruption (and hence RDI values) in biological samples.

Some patients in RCB class 3 experienced high RNA disruption at 12 weeks (Fig. 3). While this post-treatment residual disease would suggest a poor outcome, we have previously observed that patients with the highest level of mid-treatment tumour RNA disruption had post-treatment DFS durations similar to pCR recipients, whether a pCR was achieved or not [3]. This included patients with ER+/HER2- tumours, none of which achieved a pCR [3].

Merits of the addition of bevacizumab to standard chemotherapy regimens

While FEC-T + bevacizumab treatment did not result in improved survival relative to FEC-T alone [15], the former regimen was associated with higher tumour RNA disruption (Fig. 5). Perhaps there is an undefined subset of NeoAva patients that would benefit from bevacizumab treatment? The GeparQuinto clinical trial [30, 31] found that pCR rates were higher when bevacizumab was added to neoadjuvant anthracycline-taxane chemotherapy in patients with HER2- primary and metastatic breast cancer, respectively. However, the addition of bevacizumab did not increase DFS in the GeparQuinto trial [31]. Similarly, survival at 3.5 years for women with HER2- early breast cancer was not increased by bevacizumab addition in the ARTemis trial [32]. A recent meta-analysis of trials involving women with HER2- breast cancer [33] concluded that there was a significantly higher pCR rate (but no difference in disease recurrence or death) when bevacizumab was co-administered with anthracycline/taxane chemotherapy. This further emphasizes the limitations of the pCR rate as a predictor of outcome after neoadjuvant chemotherapy. Importantly, the long term outcome in patients with hormone receptor positive breast cancer is also dependent on adjuvant endocrine therapy, which would not be reflected in the pCR and/or RCB assessments post neoadjuvant chemotherapy.

Biological processes associated with low tumour RNA disruption

Preliminary exploratory RPPA studies (Supplemental Data Files 1 and 2) suggest that low tumour RNA disruption is associated with low apoptotic balance (p = 0.001), where pro-apoptotic proteins such as Bak and Bid were expressed at a significantly lower level, while pro-survival proteins such as Bcl-xl and Mcl1 were more highly expressed. Low-RDI tumours also showed significant changes in the expression of proteins associated with epithelial mesenchymal transition (EMT; p < 0.001); such changes are associated with tumour cell migration and invasion and poor patient outcome [34, 35]. Low-RDI tumours also exhibited significantly higher expression of several G1-S checkpoint proteins (p = 0.001) and DDR proteins (p = 0.001). High expression of DDR proteins could indicate activation of DNA damage repair rather than the induction of apoptosis in response to chemotherapy treatment. The increased expression of G1-S checkpoint proteins in low-RDI tumours further suggests that such tumours were in a proliferative state during treatment. While these observations provide insight into pathways involved in treatment-induced RNA disruption, it should be cautioned that, in contrast to the RDA studies involving all 132 patients in the NeoAva clinical trial, only biopsies from 77 NeoAva patients were subjected to RPPA studies. This resulted in the elimination of a large number of patients whose tumour or clinical characteristics may have impacted on the identification of pathways associated with RNA disruption. Nevertheless, the very low and significant p-values obtained for the small dataset suggests that the above pathways may be highly relevant to chemotherapy-dependent RNA disruption in vivo.

The need to validate our findings in a large independent cohort of patients with HER2- breast cancer

While the findings of this study support the potential utility of mid-treatment RNA disruption assessment to predict outcome from neoadjuvant chemotherapy, the study has several limitations. Our study assessed the potential utility of on-treatment tumour RNA disruption measurements to predict disease-free survival after neoadjuvant chemotherapy at the univariable level in a small cohort of NeoAva trial patients. Given the limited sample size of approximately 100 patients with predominantly ER+/HER2- breast cancer, no multivariable Cox analyses were performed. The retrospective study design permits potential over-fitting of data, with little control for type I and type II statistical errors. Thus, at this point, we cannot claim that the RNA disruption index is an independent marker for DFS.

Recognizing these deficiencies, we initiated a study called BREVITY (Breast Cancer Response EValuation for Individualized TherapY, clinicaltrials.gov ID NCT03524430). In this prospective study, a large independent cohort of > 450 patients with early breast cancer from 40 cancer treatment centers across 7 countries (Italy, Germany, France, Spain, Poland, Canada, and the U.S.A.) are being recruited in order to validate the utility of the RNA disruption assay to predict outcome from neoadjuvant chemotherapy in patients across tumour subtypes, employing any drug combinations currently used in breast cancer treatment, including immunomodulatory drugs and antibody-drug conjugates. Ultrasound-guided tumour biopsies are being collected 35 ± 4 days and 55 ± 5 days after the initiation of neoadjuvant chemotherapy. If the regimen changes during treatment (for example, taxane administration after anthracycline-based chemotherapy or vice versa), the second biopsy is being collected just prior to administration of the second cycle of the new regimen. RNA is then isolated from the biopsies and the degree of tumour RNA disruption assessed using the RNA disruption assay. The relationship between the maximum level of tumour RNA disruption at either of the two time points and both the pCR rate and DFS will then be assessed. The prospective study design, the previously established statistical analysis plan, a third-party clinical research organization overseeing the trial, the blinding of investigators, and the larger sample size will strongly help to minimize type I and type II errors. The larger sample size will also permit a comparison of the RDA with other markers/variables in order to assess its ability to independently predict outcome from neoadjuvant chemotherapy. Eighty BREVITY patients across the tumour subtypes were selected to train the assay. Recently published data from this BREVITY training set indicated that tumour RNA disruption is much higher in pCR recipients than in patients with residual disease (p = 0.008) [7]. Moreover, low tumour RNA disruption during treatment was found to be associated with the lack of pCR across the tumour subtypes with a NPV of 93.3% [7]. The ER+/HER2- subtype was not predominant in this training set. Thus, it would appear that the RDA need not be restricted for use only in patients with ER+/HER2- tumours. The RDA technical parameters (including RDI cut points) have now been locked down for a second cohort of > 450 BREVITY patients for validation of RDA’s ability to predict treatment outcome. Accrual is currently approximately 80% complete.

Concluding remarks and possible impact of the RDA on patient care

The lack of a consistent association between increased pCR rates and improved DFS [30–33, 36] limits the utility of pCR as a predictor of treatment outcome. In contrast, high tumour RNA disruption mid-treatment predicted for improved DFS across tumour subtypes, including ER+/HER2- tumours [3]. Consistent with this view, high tumour RNA disruption in the NeoAva study was associated with improved DFS and BCSS (Fig. 4A and B), while pCR incidence did not (Fig. 4D). This suggests that better chemotherapy outcome predictions may be possible using the RDA during neoadjuvant chemotherapy rather than waiting for post-treatment RCB, pCR, or ctDNA assessments. As suggested recently [37], on-treatment biomarkers (including possibly the RDA), may permit oncologists to identify during chemotherapy patients at high risk of treatment failure and disease progression. This could help guide treatment escalation or de-escalation decisions.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

BCSS

Breast cancer-specific survival

CI

Confidence interval

CONSORT

Consolidated standards of reporting trials

DDR

DNA damage repair

DFS

Disease-free survival

EMT

Epithelial mesenchymal transition

ER

Estrogen receptor

FEC

5-fluorouracil, epirubicin, cyclophosphamide

FEC-T

5-fluorouracil, epirubicin, cyclophosphamide followed by a taxane

HER2

Human epidermal growth factor receptor 2

M

Median

pCR

pathologic complete response

PR

Progesterone receptor

RCB

Residual cancer burden

RDA

RNA disruption assay

RDI

RNA disruption index

RNA

Ribonucleic acid

RPPA

Reverse phase protein array

rRNA

ribosomal ribonucleic acid

T

Taxane

Author contributions

O.E. is the principal investigator of the NeoAva trial. He secured trial funding, design, protocols, objectives, and biopsy/data collection. With H.V.L.G., he provided access to trial data. M.H.H. performed additional analyses with the RDA data and, along with O.C.L., H.V.L.G, and O.E., discussed analytical approaches with Rna Diagnostics. L.D. collated all data received from NeoAva investigators and performed some initial data assessments. T.M., G.T. and R.S. performed RDA analyses on the provided electropherogram data. M.A.D oversaw the RPPS pathway analyses. L.B.P assessed the collated data, prepared manuscript figures and tables, and worked with A.M.P. in writing, editing, and submitting the manuscript. All authors reviewed the manuscript and consented to its submission for publication.

Funding

This study was funded by the Norwegian Breast Cancer Society (project 11003001) and the Norwegian Research Council (project 191436/V50), with support of the K. G. Jebsen Center for Breast Cancer Research and the South-Eastern Norway Regional Health Authority (projects 2020072 and 2024099, respectively). The NeoAva trial was supported by Roche Norway and Sanofi-Aventis Norway.

Data availability

The raw data supporting the results reported in this article can be found in Supplemental Data Files 1, 2, and 3. Access to NeoAva materials and data is governed by its principal investigator Dr. Olav Engebraaten.

Declarations

Ethics approval and consent to participate

The NeoAva trial was approved by the Institutional Protocol Review Board, regional research ethics committees, and the Norwegian Medicines Agency. The trial was conducted in accordance with the Declaration of Helsinki International Conference on Harmonization Good Clinical Practice. All patients provided informed consent prior to inclusion in the NeoAva clinical trial.

Consent for publication

All patients agreed to the sharing and publication of pseudonymized data. All authors reviewed the manuscript and consented to its submission for publication.

Competing interests

Some authors declare significant conflicts of interest, which did not impact on study design or interpretation. L.B.P., T.M., R.S., G.T. and L.D. are/were employees of Rna Diagnostics, which seeks to introduce the RDA into the oncology marketplace. A.M.P. and L.B.P. are minority shareholders in the company. A.M.P. receives research support from Rna Diagnostics. O.E. received research support from Roche Norway and Sanofi-Aventis, who helped fund the NeoAva trial. The remaining authors declare no conflicts of interest.