Preoperative β-Blockade with Propranolol Reduces Biomarkers of Metastasis in Breast Cancer: A Phase II Randomized Trial

This article is featured in Highlights of This Issue, p. 1779

Metastasis is a critical stage of breast cancer progression associated with a dramatic reduction in life expectancy. During times of stress, activation of the sympathetic nervous system (SNS) releases neurotransmitters that act at β-adrenergic receptors on tumor cells and tumor-associated immune cells to promote metastasis (1–3). Prior to cancer surgery, psychobiologic processes (e.g., stress of cancer diagnosis) and physiologic processes (e.g., surgical stress) converge to increase SNS activity. The resulting increase in β-adrenergic signaling may increase vulnerability to initiation and promotion of metastasis (4, 5).

To the extent that perioperative SNS neurotransmitters act through β-adrenergic receptors on residual or micrometastatic disease to promote cancer recurrence, β-adrenergic antagonists, or β-blocker drugs, may abrogate this risk (5, 6). Consistent with this hypothesis, observational pharmacoepidemiologic studies have indicated a protective effect of incidental β-blocker exposure on disease-free or overall survival in breast cancer (7–10), although results are mixed (11). In vivo studies using mouse models of cancer have identified plausible mechanisms, finding that β-blocker treatment with propranolol prevents breast cancer invasion (12), inflammation and immune suppression (1, 3, 13), and metastasis (1, 2, 14).

To investigate whether β-blockade has antimetastatic properties in patients, we conducted a phase II randomized placebo-controlled trial to test the hypothesis that preoperative β-blockade with the nonselective drug propranolol reduces expression of prometastatic and proinflammatory genes in the tumor microenvironment. The primary endpoint of this study was the quantitative change in prometastatic and proinflammatory gene expression within the tumor between baseline (analyzing the diagnostic biopsy, obtained prior to treatment) and surgical resection (after 7 days of β-blocker treatment). Secondary analyses assessed the clinical effects of propranolol on preoperative anxiety and perioperative cardiovascular parameters.

This prospectively registered (ACTRN12615000889550), single-center, triple-blinded placebo-controlled randomized clinical trial was conducted at Peter MacCallum Cancer Centre (Melbourne, Australia). Patients were recruited from January 2016 to September 2017. Ethics approval was obtained from the Institutional Review Board (Peter MacCallum Cancer Centre Human Research Ethics Committee, July 9, 2015; HREC 14-139). The study was conducted in accordance with the ethical guidelines of the Declaration of Helsinki, and written informed consent was obtained from all patients before performing any study-related procedures.

English speaking female patients ages 18 to 80 years (Eastern Co-operative Oncology Group <2) with a diagnosis of surgically resectable primary breast cancer were eligible for this study. Patients were seen by a clinically independent study investigator in the breast surgical outpatient clinic and invited to participate in the study. Once written informed consent was achieved, patients were registered as study participants. Exclusion criteria included patients who were pregnant, breastfeeding, receiving repeat breast cancer surgery (within 6 months), receiving neoadjuvant chemotherapy or hormonal therapy, had a histologic diagnosis of isolated ductal carcinoma in situ, had contraindications to propranolol, were recently or regularly using medication including β-blockers (within the last 3 months), nonsteroidal antiinflammatory drugs (daily use), anxiolytics, calcium channel blockers, or α-adrenergic receptor agonists, or had a history of moderate or severe asthma, or stroke.

Sixty-four patients were recruited into this trial; recruitment continued until 60 patients had completed treatment per protocol. Patients were asked to provide demographic data, a medical and socioeconomic history, and baseline observation data were recorded. Patients were randomly assigned in a 1:1 ratio to receive either placebo or propranolol for 7 days prior to their surgery date. The study was triple-blinded: the participant, treating physician, and the investigators examining the primary endpoint were blinded to treatment allocation (15).

Patients were instructed to commence the study drug (40 mg oral propranolol or placebo) twice a day starting 7 days prior to the date of surgery. Patients were provided with a dosing schedule, a home monitor (Omron Blood Pressure Monitor – HEM7130; Omron Healthcare) for self-assessment of blood pressure and heart rate, and a chart for daily recording of symptoms and hemodynamic data. After 3 days of treatment, the dose of study drug was escalated to 80 mg oral propranolol or placebo a day daily until the day of surgery (Fig. 1A) if the following three predetermined criteria were met: systolic blood pressure was greater than 110 mm Hg, and heart rate was greater than 65 beats per minute (bpm), and they had no significant symptoms (e.g., syncope, insomnia, fatigue). After surgery, patients were weaned from study medication over 3 days (Fig. 1A). All investigator-assessed measurements of blood pressure and heart rate occurred at the institution in which the trial was conducted. Study drug consumption was verified by patient self-reporting (log-book), supervised administration during hospital admission, and tablet count in returned study drug vials. For cancer surgery, all patients were prescribed a protocolized anesthetic (Supplementary Materials and Methods), and intraoperative hemodynamic data were collected. On the day of surgery, subjective patient anxiety was assessed using a 3-point Likert scale. Patients were asked whether their anxiety and stress levels were “better,” “no change,” or “worse” than prior to commencing drug treatment.

The sample size calculation for gene expression profiling and bioinformatic analysis is described in the Supplementary Materials and Methods. Total RNA was extracted from 5 μm sections of formalin-fixed paraffin-embedded (FFPE) core biopsy tissue (baseline, obtained at diagnosis) and the surgically resected primary tumor (outcome; obtained at surgical excision of the primary tumor; Qiagen RNeasy FFPE), RNA was tested for suitable mass (RiboGreen) and integrity (Agilent TapeStation), reverse transcribed to complementary DNA (Lexogen QuantSeq 3′ FWD), and sequenced on a HiSeq 4000 instrument (Illumina) in the UCLA Neuroscience Genomics Core laboratory, following the manufacturers' standard protocols. Sequencing targeted >10 million 65-nt single-stranded reads per sample (achieved mean 13.5 million/sample), which were mapped to the human transcriptome and quantified as transcripts per million mapped reads using the STAR aligner. Transcript abundance values were log₂-transformed for statistical analysis of change over time (treated tumor minus baseline diagnostic biopsy), using standard linear statistical models to quantify the magnitude of differential change over time in propranolol-treated patients relative to placebo-treated controls. A priori hypotheses regarding epithelial–mesenchymal transition (EMT) polarization and tumor-associated leukocyte transcriptomes were tested using Transcript Origin Analysis (16). All genes found to show ≥1.25-fold differential change over time in propranolol versus placebo samples were mapped to cell-type diagnosticity scores derived from previous reference studies of transcriptome profiling of mesenchymal versus epithelial-polarized breast cancer cells (GSE13915; ref. 17), isolated leukocyte subsets (CD4⁺ and CD8⁺ T cells, B cells, NK cells, classical and nonclassical monocyte subsets, and DC1, DC2, and DC3 dendritic cell subsets; GSE101489; ref. 18), or isolated M1- and M2-polarized macrophages (GSE5099; ref. 19).

A priori hypotheses regarding activity of EMT-relevant (Snail/Slug, Smad) and proinflammatory (NF-κB/Rel, AP-1) transcription control pathways were tested using TELiS bioinformatic analysis (20) of transcription factor–binding motifs (TFBM) in the promoters of all genes showing ≥1.25-fold differential expression. TELiS uses a z-test to identify TFBM enrichment ratios significantly greater than one (or significantly less than one). In all bioinformatics analyses, standard errors (SEs) were estimated by 200 cycles of bootstrap resampling of observed linear model residual vectors to quantify the effect of sampling variability on the set of genes identified as differentially expressed, which controls for association among genes.

The effect of propranolol on primary tumor immune status was measured by iIHC on full-face tumor sections, performed on a Dako Autostainer Link 48 (Agilent) and a Ventana BenchMark Ultra (Roche Ventana). Positive and negative human control tissues were used to assess staining for both individual slides and split runs, and a section from each tissue block was stained by hematoxylin and eosin. Full protocol and antibody details are provided (see Supplementary Table). The extent of immunostaining was independently scored: 0, “no reactivity”; 1, “sparse”; 2, “mild”; 3, “dense”; 4, “very dense.” For analysis, the extent of immune cell infiltrate was dichotomized on the basis of low (1–2) or high intensity (2–4).

Blood samples were taken at baseline and at surgery to analyze the effect of propranolol on peripheral inflammatory cytokines (see Supplementary Materials and Methods). Serum levels of propranolol on the day of surgery were determined using an Acquity UPLC System (Waters) coupled to a Micromass Quattro Premier MS (Waters; positive mode electrospray ionization). All samples were analyzed in a single run with quantitation relative to calibration standards prepared in blank human serum. Assay performance was verified over the concentration range of 1 to 2,000 ng/mL, which adequately encompassed all patient samples.

Continuous data were tested for normality using the Kolmogorov–Smirnov test. Normally distributed data were presented as mean (SD) and compared using Student unpaired two-tailed t tests. Nonparametric data were presented as median (interquartile range) and compared using Mann–Whitney U tests. Categorical data were summarized as frequency (percentage) and compared using a χ² test or Fisher exact test as appropriate. A value of P < 0.05 was considered statistically significant. The primary endpoint investigated quantitative change in tumor tissue prometastatic gene expression from baseline biopsy to resected primary tumor and was analyzed by both intention-to-treat and per protocol principles; all other analyses were conducted per protocol. Bioinformatics parameter estimation and bootstrap resampling were conducted using the Java JAMA matrix algebra package; all other statistical analyses were conducted using SPSS for Windows (version 23; IBM). The sample size of 60 was selected by power analysis based on effect size estimates from previous research (detailed in Supplementary Materials and Methods).

In this randomized placebo-controlled triple-blind phase II study of perioperative propranolol versus placebo in patients with breast cancer (Fig. 1A), 64 patients were recruited, registered, and randomly assigned to treatment groups (Fig. 1B). Four patients were omitted from per-protocol analysis: 3 patients self-withdrew due to a change in health condition prior to commencing study medication, and 1 patient's surgery date was made urgent and per-protocol dosing prior to surgery was not feasible. Baseline demographic data and disease status were comparable between groups (Table 1).

Compliance with the 15 preoperative dosing episodes was high (96%); 2 patients incorrectly omitted the dose on the morning of surgery. During the preoperative dosing phase, 15 patients (50%) were not dose-increased above 40 mg twice a day due to: not achieving blood pressure criteria (propranolol: 12, placebo: 0); not achieving heart rate criteria (propranolol: 1, placebo: 1); fatigue (propranolol: 1, placebo: 1); or insomnia (propranolol: 1, placebo: 0). The serum concentration of propranolol on the day of surgery varied by the propranolol dose taken that morning: 0 mg (n = 2, median 7.5 ng/mL, IQR = 5.8–9.1), 40 mg (n = 13, median 18.0 ng/mL, IQR = 11.2–32.1), and 80 mg (n = 15, median 65.6 ng/mL, IQR = 54.6–110.0).

Overall, study drug treatment was well tolerated. One patient in each group experienced asymptomatic bradycardia (heart rate <50 bpm) and 3 patients in the propranolol group experienced asymptomatic hypotension (systolic blood pressure <100 mmHg); no episodes of bronchospasm were recorded. Propranolol lowered preoperative heart rate and systolic blood pressure (Fig. 2A and B), and intraoperative heart rate, but not intraoperative systolic blood pressure (Supplementary Fig. S1). On the morning of surgery after a week of preoperative dosing, propranolol-treated patients reported a reduction in anxiety, compared with placebo-treated patients (P < 0.025), but no significant reduction in stress (P = 0.12; Fig. 2C).

Comparative genome-wide transcription profiling of the resected primary tumors with the respective diagnostic biopsies identified 1,054 genes with ≥1.25-fold upregulation in the propranolol-treated patients relative to the placebo-treated controls, whereas 59 genes showed equivalent downregulation.

To determine whether the observed differences in gene expression reflected an alteration in tumor EMT differentiation, we conducted a priori-defined bioinformatic analyses mapping differentially expressed genes to previously derived reference profiles of epithelial- and mesenchymal-polarized breast cancer cells (GSE13915; ref. 17). Genes downregulated in tumors from propranolol-treated patients vs. controls were predominately mesenchymal in origin (diagnosticity z-score: mean 0.34 ± 0.12; P = 0.002; Fig. 3A). Genes upregulated by propranolol trended toward epithelial origin but this difference did not reach statistical significance (0.07 ± 0.09, P = 0.21).

As a convergent validation of primary results, we conducted additional bioinformatics analyses of gene regulation pathways known to modulate EMT, based on asymmetries in the prevalence of transcription-factor binding motifs (TFBM) in the promoters of upregulated versus downregulated genes. We tested two a priori hypotheses. The first examined whether the EMT-driving Snail/Slug and Smad transcription factors were downregulated by propranolol. Results indicated a significant reduction in Snail/Slug activity in tumors from propranolol-treated patients (−0.43 log₂ ratio in upregulated vs. downregulated gene promoters, ±SE 0.19, P = 0.03; Fig. 3B) and a trend toward reduced Smad activity (−0.19 ± 0.13, P = 0.11). The second analysis tested for downregulated activity of inflammation-related NF-κB/Rel and AP-1 transcription factors, and found both to be downregulated by propranolol (NF-κB/Rel: −1.06 ± 0.41, P = 0.01; AP-1: −0.80 ± 0.19, P < 0.01; Fig. 3B). Catecholamine signaling through β-adrenergic receptors modulates gene expression through the cAMP response element (CRE) recognized by the CREB transcription factor. Therefore, we tested for a potential effect of propranolol on signaling through the CREB pathway and found significant reduction in its activity within the primary tumor (−0.48 ± 0.23, P = 0.04).

To characterize the effect of β-blockade on immunologic status of tumors, we conducted Transcript Origin Analysis using profiles of isolated leukocyte subpopulations as references (GSE101489; refs. 16, 18). These analyses found propranolol downregulated genes within tumors that are characteristic of neutrophils (diagnosticity score: −2.54 ± 0.48, P < 0.0001), and to a lesser extent, NK cells (−0.55 ± 0.13, P < 0.0001; Fig. 3C). Upregulated genes derived from classical myeloid DC1/BDCA1+ dendritic cells (0.17 ± 0.04, P < 0.0001) and DC2/BDCA2+ plasmacytoid dendritic cells (0.18 ± 0.08, P = 0.02), both key antigen-presenting cells (Fig. 3C). In additional analyses using M1 and M2 macrophage profiles as reference points (GSE5099; ref. 19), tumors from propranolol-treated patients showed increased M1 macrophage polarization (0.27 ± 0.12, P = 0.01; Fig. 3D).

The significant effects of propranolol on gene expression were similar regardless of whether analyses were conducted on the basis of intention-to-treat, per-protocol, or after adjustment for variations in serum propranolol concentration on the day of surgery. All further analyses were conducted per protocol. Propranolol treatment had no effect on basal plasma cytokine levels between pretreatment baseline and the immediate presurgical blood draw (P > 0.05).

Post hoc analysis identified a subset of propranolol-treated patients who did not achieve clinical signs of β-blockade despite escalation of the propranolol dose. These patients showed minimal reduction in blood pressure (less than 20 mmHg) and heart rate (less than 10 bpm) between the start of propranolol dosing (baseline) and the day of surgery (following 7 days of treatment; Fig. 4A). Dichotomization of propranolol-treated patients as clinically “β-blocked” (n = 15 of 30) or clinically “non–β-blocked” (n = 15 of 30) demonstrated a differential intratumoral immune cell response (Fig. 4B). Although propranolol modulated expression of leukocyte genes across the entire cohort (Fig. 3C), patients who were clinically β-blocked had consistently high infiltration of CD68⁺ macrophages (n = 15 of 15) and CD8⁺ T cells (n = 15 of 15), whereas immune cell infiltration was reduced in non–β-blocked patients: CD68⁺ (n = 9 of 15 non–β-blocked patients, P < 0.02, Fisher exact test) and CD8⁺ (n = 10 of 15 non–β-blocked patients, P < 0.04, Fisher exact test; Fig. 4C).

Surgery is integral to the treatment of early-stage breast cancer. However, the increase in SNS activity and β-adrenergic signaling that occurs in the perioperative period may increase the invasive potential of residual or micrometastatic disease, and drive cancer recurrence (5). This prospective, randomized placebo-controlled phase II biomarker trial demonstrates that intervening on this risk through preoperative β-blocker administration can impact multiple biological processes with the potential to reduce cancer recurrence.

In transcriptome profiling of primary breast tumors resected after 7 days of propranolol exposure, pathway analysis of empirical transcriptome alterations indicated reduced EMT polarization, reduced activity of the EMT-associated Snail/Slug transcription factor family, reduced activity of the proinflammatory NF-κB and AP-1 transcription factor families. Consistent with these findings, the cell-based bioinformatic analyses linked propranolol to reduced neutrophil infiltration/activation, increased M1 macrophage polarization, and increased recruitment/activation of BDCA1⁺ classical myeloid dendritic cells and BDCA2⁺ plasmacytoid dendritic cells, suggesting a shift in the tumor microenvironment from inflammation towards mature myeloid populations that may support anticancer immunity. Overall, this alteration in the profile of EMT, inflammatory, and immune cell dynamics is broadly consistent with enhanced cellular immune response and reduced malignant potential. As such, these biomarker findings provide a strong biological rationale for future research in large, adequately powered, patient cohorts to assess the impact of β-blockade on clinical cancer outcomes such as disease-free survival and overall survival.

The findings provide the first prospective clinical evidence that cellular and molecular mechanisms identified for β-blockers in mouse cancer models may impact cancer progression processes in patients. In preclinical cancer models, propranolol prevents EMT, reduces invadopodia formation, tumor cell invasion, and metastasis of cancer cells (1, 12, 21, 22), and enhances antitumor immunity (23–26). Here we show that single-agent propranolol is sufficient to induce similar protective changes in the tumor microenvironment in patients with breast cancer, identifying plausible mechanisms for a growing collection of epidemiologic findings that β-blocker use for other indications is associated with improved outcomes in patients with subsequent breast cancer (8–10). The findings underscore the contributions of β-blockade to changes in tumor immune status and invasion profiles that were observed in patients treated with a combination of nonselective antiinflammatory drug and the nonselective β-blocker propranolol (27). Studies in mouse cancer models and retrospective clinical cohorts suggest β-blockers also protect against progression in diverse tumor types, raising the possibility that findings of this study could be broadly applicable to other cancer types. Consistent with this hypothesis, a prospective unblinded cohort study reported that propranolol (compared with no treatment) reduced melanoma recurrence rates (HR = 0.18; 95% Confidence Interval, 0.04–0.89; P = 0.03; ref. 28).

Currently, there are no registered prospective trials of β-blockers in women with breast cancer that are powered for clinical outcomes. Findings from this study identify several points of consideration for design of future clinical studies, including the observation that even moderate doses of propranolol were sufficient to reduce perioperative anxiety and stress, and modulate progression-related gene expression in tumors, with favorable effects on biomarkers of invasion and inflammation regardless of individual dose (80–160 mg/day). Further, the findings suggest that clinical endpoints such as heart rate and blood pressure changes may provide noninvasive surrogate measures of treatment response at the level of the tumor. In contrast, serum propranolol levels on the day of surgery were not associated with outcome. In future studies it will be important to determine whether propranolol-induced changes in tumor gene expression—that here were observed regardless of the level of clinical β-blockade (Fig. 3)—more consistently lead to changes in immune cell recruitment when drug exposure is continued for longer than a week. It will also be important to determine if cellular immune recruitment induced by β-blockade predicts long-term improvement in cancer outcome, for example, by enhancing treatment outcomes with standard-of-care therapies such as immunotherapy or chemotherapy.

Future research using larger samples will be required to determine if propranolol has differential effects as a function of tumor subtype or patient characteristics. The present sample was not large enough to determine how propranolol administration might impact tumor biology differently depending on patient demographic characteristics or tumor pathology (e.g., ER/PR/Her2 status, tumor grade, etc.). Nonetheless, these factors were well-balanced across patients randomized to propranolol versus placebo, and thus do not confound the overall effect of propranolol in this study. Finally, to maximize success of future trials, careful consideration of β-blocker dose-escalation regimens is vital, as evidenced by 33% hypotension rate in a feasibility study of propranolol in patients with multiple myeloma (29). The optimal approach may depend on the fitness levels of the cohort population. In our study of early-stage breast cancer, the 12 patients who were not dose-escalated had a baseline normotensive state with (enrolment) heart rate and blood pressure that were only marginally above the preestablished cut-off values for drug escalation. Given there is no medical indication for excluding ostensibly healthy patients from propranolol treatment, future trials may consider utilizing percentage reduction from baseline, rather than absolute cutoffs, as the escalation criterion.

This trial is the first prospective, blinded, randomized clinical trial examining the effect of single-agent nonselective β-blockade with propranolol on cancer gene expression. Preoperative propranolol downregulated biomarkers of invasive potential and inflammation, and improved cellular immune response in breast cancer. Propranolol was well-tolerated by patients, appears safe, and achieved beneficial anxiolytic effects. Although trials powered for a primary endpoint of clinical outcome will be required before propranolol can be considered in breast cancer treatment, this study's biological results support the potential utility of β-adrenergic blockade as adjunctive therapy to control minimal residual disease in women with breast cancer.

E.K. Sloan is an unpaid consultant/advisory board member for Cygnal Therapeutics. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J.G. Hiller, S.W. Cole, B. Riedel, E.K. Sloan

Development of methodology: J.G. Hiller, S.W. Cole, M.A. Henderson, P.S. Myles, S. Fox, B. Riedel, E.K. Sloan

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.G. Hiller, E.M. Crone, D.J. Byrne, D.M. Shackleford, J.-M.B. Pang, M.A. Henderson, S.S. Nightingale, S. Fox, E.K. Sloan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.G. Hiller, S.W. Cole, D.M. Shackleford, K.M. Ho, S. Fox, E.K. Sloan

Writing, review, and/or revision of the manuscript: J.G. Hiller, S.W. Cole, D.J. Byrne, D.M. Shackleford, M.A. Henderson, S.S. Nightingale, K.M. Ho, P.S. Myles, B. Riedel, E.K. Sloan

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.G. Hiller, S.W. Cole, E.M. Crone, D.J. Byrne, E.K. Sloan

Study supervision: J.G. Hiller, M.A. Henderson, P.S. Myles, B. Riedel, E.K. Sloan

The authors are grateful to Drs David Gyorki, Cathie Poliness, and David Speakman for assistance with patient recruitment; Dr Alexander Ziegler for assistance with soluble cytokine analysis; and Aeson Chang for constructive discussion. The work of the authors was supported by the NCI Contract No. HHSN261200800001E, and the NCI Network on Biobehavioral Pathways in Cancer, the Australian and New Zealand College of Anaesthetists (project grant 16/004), National Health and Medical Research Council 1147498, the National Institute of Aging P30 AG017265, Perpetual Trustees IMPACT Philanthropic 2016 Research Grant, and The David and Lorelle Skewes Foundation. P.S. Myles was supported by an NHMRC Practitioner Fellowship (ID1135937).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.