Development of a Breast-on-a-Chip Microfluidic Model to Assess the Effect of Palbociclib in MCF-7 and T47D Cancer Cells

Highlights

What are the main findings?

• A controlled perfusion (20 µL/h) was established, allowing simultaneous evaluation of proliferation, viability, cytoskeletal organization, apoptosis, and PARP1 differential processing in luminal breast cancer cells under flow conditions.
• Palbociclib treatment produced dose-dependent changes in cell viability, morphology, apoptotic response, and PARP1 processing in MCF-7 and T47D cells, with comparable biological response trends observed across distinct microfluidic setups operated under identical flow and temperature parameters.

What are the implications of the main findings?

• The well-controlled 2D microfluidic breast-on-a-chip system provides a robust and experimentally accessible framework for studying drug response and resistance-associated biomarkers under physiologically relevant dynamic conditions.
• This approach represents a scalable starting point for organ-on-a-chip applications and supports future integration of limited cell sources, including patient-derived samples, to improve translational relevance in anticancer drug screening.

Abstract

Organ-on-a-chip devices combine microfabrication, tissue engineering, and microfluidics to recreate physiologically relevant microenvironments for in vitro studies. In this work, we validated a dynamic 2D breast-on-a-chip microfluidic bioassay operated at a controlled infusion rate of 20 µL/h to assess anticancer drug responses under defined flow conditions. Using Palbociclib as a reference compound, we evaluated proliferation, viability/apoptosis, cytoskeleton organization, and differential processing of the resistance-associated marker PARP1 in MCF-7 and T47D breast cancer cells. Under dynamic microfluidic conditions, Palbociclib induced dose-dependent effects, with the higher concentration (20 µM) consistently reducing cell proliferation and viability and increasing late apoptosis compared to 10 µM Palbociclib. Cytoskeletal disorganization was observed at both concentrations, while differential PARP1 processing patterns between MCF-7 and T47D cells were detected across doses. These responses are consistent with known effects of CDK4/6 inhibition and were reproducibly captured under controlled flow conditions. Overall, our results demonstrate that this breast-on-a-chip microfluidic model provides a reproducible and physiologically relevant in vitro platform for integrated assessment of drug efficacy and resistance-associated markers under dynamic perfusion.

1. Introduction

Organs-on-a-chip (OoC) are typically plastic chips containing cells, tissues, or organoids/mini-organs that are used to simulate the microenvironment and micro-physiological conditions of tissues and organs on a smaller scale, aiming to test new medications and treatments for health disorders. OoC are widely used to simulate physiological and pathological tissue conditions and structures, as well as their microenvironmental components (e.g., blood, gas, force, etc.) [1,2]. OoC are developed using technologies such as microfabrication, microfluidics, and bioprinting, and are applied for cell/organ/tissue culture, maintaining their structure and dynamic microenvironments on small devices. In addition, OoC can be integrated with sensor systems to achieve continuous and automatic detection of biochemical and physical parameters [2,3,4]. OoC platforms typically contain chips with tissues arranged in two- or three-dimensional formats (2D–3D), such as spheroids, which promote cell–cell and cell–tissue interactions and better reflect physiological conditions within a controlled microenvironment [3]. These devices can more closely mimic human physiology by integrating biomechanical stimuli, such as tensile and compressive forces present in lung and muscle tissues, as well as the hemodynamic shear stress characteristic of vascular environments. Hence, variations in how these biomechanical forces are simulated inside the chips are known to influence tissue inflammation and drug absorption [2,5,6,7]. The development of OoC systems is challenging due to the complexity of mimicking large-scale physiological tissues and organ-level functions within miniaturized models. OoC maintains the main characteristics of human tissues and provides relevant knowledge about disease mechanisms, supporting the development of effective treatments [8,9,10,11,12].

Most drug effect assessment models are in vitro 2D and 3D cell cultures or animal models. The conventional in vitro models usually include culture flasks, transwell plates, scaffolds, or 3D spheroids, reproducing the tumor microenvironment. The 3D models can be specifically used to evaluate cellular responses to treatment, such as viability, proliferation, migration, and drug resistance [13,14]. These non-fluidic in vitro models are not capable of reproducing the fluidic forces of the bloodstream, tumor microenvironment, tissue deformation, and shear stress, which play a major role in cancer cell invasion [13,15]. Despite animal models being able to better mimic the biological and structural complexities of the tumor microenvironment, they cannot be used to estimate the real effect of drugs in humans [13,16].

Considering that 2D cell culture in vitro models do not provide the necessary tissue microenvironment for drug testing and animal models differ from human models, OoC were created to promote a 3D organ development closer to human models. At first, 3D organoid culture techniques were developed and later evolved into organ-on-a-chip systems based on microfluidic technology. Organoid culture typically involves the 3D growth of mammalian stem cells with the sequential addition of growth factors; however, it remains difficult to precisely control the conditions required to reproduce the complex and dynamic microenvironment of a whole organ. In contrast, OoC systems enable an accurate regulation of the microenvironment within the microfluidic chip, and organoids can be integrated into these platforms to generate more complex and physiologically relevant models. In recent years, microfluidics-based artificial in vitro organs have been developed for multiple organ types, including breast, lung, liver, heart, and kidney, as well as for distinct models of the same organ [17]. Breast-on-a-chip models have been used in recent years to more comprehensively understand disease mechanisms and evaluate new therapies for breast cancer treatment. To simulate a realistic breast structure and function, stem cells or human primary cells have been commonly used in the development of spheroids and luminal or other functional units through differentiation, self-organization, or 3D bioprinting. More recently, breast cancer metastasis models were built connected to other organs (like multi-organs-on-a-chip) to evaluate metastasis mechanisms or anti-metastatic drug testing, such as breast-liver-on-chip, breast-vessel-on-chip, breast-heart-on-chip [17].

Although Organ-on-a-Chip platforms are often associated with three-dimensional tissue architectures, physiological relevance depends not only on dimensionality but also on the presence of dynamic vascular-like conditions. For anticancer drug testing, controlled perfusion, continuous drug delivery, shear stress, and efficient metabolic waste removal are critical determinants of cellular responses. In this context, 2D microfluidic systems enable precise control of flow, homogeneous drug exposure, and defined mechanical stimuli that can be difficult to achieve in many static 3D cultures. The proposed 2D breast-on-a-chip model provides sustained perfusion at defined infusion rates and capillary-level shear stress, allowing reproducible interrogation of drug efficacy and resistance-associated responses under dynamic conditions. Rather than replacing 3D organoid or spheroid models, which are widely regarded as physiologically relevant in many contexts, this approach is intended to complement them by providing a streamlined and experimentally controllable platform, particularly suited for mechanistic studies and applications with limited cell availability, such as patient-derived samples.

Many breast-on-a-chip models bearing different customized features have been used for different purposes, such as the screening of new drugs, the evaluation of antitumor drugs and chemotherapies’ effect, cytotoxicity and viability assays, the study of tumor microenvironment, the study of tumor invasion and angiogenesis, and the evaluation of immunotherapies directing patients’ own T-cells to target and kill cancer cells. Some metastasis-on-chip models have employed triple-negative breast cancer cells for metastasis studies, paracrine signaling, cytokine expression analysis, and cancer aggressiveness in microfluidic chips [18]. Different breast cancer cells (e.g., triple negative breast lineages, MCF-7, MDA-MB-453, MDA-MB-231, and HCC1937) have been used to assess the effect of PARP1 inhibitors, Palbociclib, and similar drugs (CDK inhibitors) on tumor cell viability, proliferation, and integrity [19,20]. Other breast-on-a-chip microfluidic models have been used to evaluate other drugs’ cytotoxicity, their effect on breast cells’ viability [21], monocytes polarization into tumor-associated macrophages (TAMs) and tumor cell migration [22]. A recent breast cancer-on-chip model with an integrated endothelial barrier that enables immune cells infiltration into the tumor and cytokine release has been developed to study patient-specific CAR-T cell efficacy [23]. Therefore, these BoC models not only favor animal replacement but, more importantly, enable the customization of treatments based on the patient’s own tumor sensitivity and resistance, paving the way for personalized medicine.

Palbociclib is an antitumor anti-proliferative drug that inhibits CDK4/CDK6 kinases [24] of the proliferation signaling pathway used for the treatment of estrogen receptor-positive and human epidermal growth factor receptor 2-negative (ER+/HER2−) breast cancer [25]. Palbociclib inhibits the phosphorylation of the retinoblastoma protein (Rb), thus retaining cells in the G1-phase of the cell cycle [26]. The low-phosphorylated Rb protein inhibits the transcription factor E2F, thereby hindering the transcription of genes related to S-phase and proliferation. Other types of cancer that retain Rb expression and depend on the CDK4/6 pathway for proliferation are expected to be affected by CDK4/6 inhibitors such as Palbociclib. However, CDK4/6 inhibitors have more impact on ER+/HER2− breast cancers, especially because the proliferation pathway dependent on ER requires the activation of the CDK/Rb/E2F signaling pathway [25,26]. In addition, persistent mTORC1 signaling during Palbociclib-induced G1 arrest is a potential liability for ER+ breast cancer cells [27]. In estrogen receptor-positive (ER+) breast cancer cell lines, Palbociclib induces phenotypic alterations related to senescence [28], which can lead to cytoskeleton disturbance and cell death. Hence, in this work, we used Palbociclib for the standardization of a new microfluidic platform for antitumor drug screening/testing.

It is known that PARP1 is cleaved into two fragments, one of 89 kDa and the other of 24 kDa. After PARP1 cleavage by caspases, the 24 kDa PARP1 fragment irreversibly binds to DNA breaks and acts as a transdominant inhibitor of active PARP1, whereas the 89 kDa PARP1 fragment is translocated to the cytoplasm [29,30]. Consequently, PARP1 fragmentation by caspases leads to its inactivation, which inhibits DNA repair and facilitates caspase-mediated DNA fragmentation in apoptosis. Following their synthesis by PARP1 in the nucleus, poly(ADP-ribose) (PAR) polymers translocate to the cytoplasm, where they bind to apoptosis-inducing factor (AIF), a protein normally anchored to the mitochondrial membrane [31,32]. PAR binding to AIF results in its release to the cytoplasm [33]. As AIF has a nuclear localization sequence (NLS) near its N terminus, released AIF is translocated to the nucleus and associates with DNAase, resulting in large-scale DNA fragmentation [34]. PAR also interacts with hexokinase 1, which is the first enzyme in the glycolytic pathway, and inhibits its activity, leading to energy depletion [35]. These pathways, including PAR synthesis and its translocation to induce programmed cell death, are caspase-independent [36].

It is known that apoptosis inducers staurosporine and actinomycin D promote caspase-3 activation, and induction of PARP1-mediated PAR production, and then PARP1 cleavage into 89 kDa and 24 kDa PARP1 fragments [37]. According to Mashimo and colleagues (2020), the 89 kDa PARP1 cleaved form, with attached PAR polymers, is translocated from the nucleus to the cytoplasm, whereas 24 kDa fragments are associated with DNA breaks [37]. The cleaved 89 kDa PARP1 form in the cytoplasm interacts with AIF via their PAR polymers, inducing AIF release and translocation to the nucleus, resulting in nuclear shrinkage [37]. Thus, the cleaved 89 kDa PARP1 fragment generated by caspase-3 acts as a PAR carrier from the nucleus to the cytoplasm to promote AIF-mediated DNA fragmentation in caspase-mediated apoptosis [37]. In this work, the presence of all three forms of PARP1, 116 kDa, 89 kDa, and 24 kDa, was evaluated.

This work aimed to standardize and validate a new bench-top microfluidic platform developed by our group for breast-on-a-chip–based drug-response assessment under continuous perfusion. Validation was performed using the well-established CDK4/6 inhibitor Palbociclib to evaluate treatment responses of MCF-7 and T47D breast cancer cells, including proliferation, viability, cytoskeletal organization, and PARP1 processing as a resistance-associated marker. To ensure robust validation under controlled conditions, the breast-on-a-chip model was operated in parallel using a commercial syringe-pump system housed in a 37 °C incubator and the newly developed bench-top platform incorporating an integrated heating chamber, both configured with identical temperature and infusion-rate parameters. Importantly, this study was not designed to compare microfluidic systems or culture dimensionality (2D ×3D), but rather to validate the bench-top platform against a commercial reference using standardized operational conditions. Comparable cellular responses were observed across both microfluidic setups, confirming the reliability and reproducibility of the new platform. Thus, the focus of this work is platform validation and robustness under physiologically relevant dynamic parameters, rather than the identification of novel biological effects of Palbociclib.

2. Materials and Methods

2.1. Microfluidic Setup for the Breast-on-a-Chip System

The 2-chamber model 844 chips (2 × 10⁵ cells/chamber; cat. 10001002, ChipShop, Jena, Germany) and the 4-chamber model 221 chips (1 × 10⁵ cells/chamber; cat. 10000293, ChipShop) were used to culture MCF-7 and T47D breast cancer cells and evaluate the effects of Palbociclib under controlled microfluidic conditions (Figure 1A,B). Two microfluidic platforms were employed: (i) a standard syringe-pump system (Insight Equipamentos, Ribeirão Preto, Brazil, cat. EFF 311) operated inside a 37 °C incubator with the connected chips (Figure 1C,D), and (ii) a custom bench-top platform developed by our group, equipped with an integrated temperature-controlled chamber maintained at 37 °C, eliminating the need for an external incubator (Figure 1E–G). Both platforms were operated as closed systems without gas exchange, and CO₂ supplementation was not required for pH control (pH was measured prior to and after treatment). To ensure a high level of standardization and reproducibility between platforms and between chip types, all experiments were performed under identical and strictly controlled conditions. Prior to biological assays, both systems were systematically validated to confirm accurate flow control and the precise volume delivered through the microfluidic circuit over time. Identical 5 mL syringes were used in both platforms, and a total volume of 1.5 mL of culture medium containing Palbociclib was infused at a constant flow rate of 20 µL/h for 48 h. This flow rate was selected based on preliminary optimization experiments showing that rates below 10 µL/h reduced drug absorption, whereas rates above 20 µL/h caused cell detachment. Hence, both microfluidic configurations were operated at an identical infusion rate (20 µL/h) and controlled temperature (37 °C). This approach ensured standardized dynamic perfusion parameters across experimental setups.

The 4-chamber chip 221 allowed the simultaneous incubation of all experimental groups (Flux+, Flux−, Static+, and Static−) within a single chip; therefore, it was incubated inside the temperature-controlled chamber of the bench-top platform, minimizing inter-chip and inter-platform variability. Chip 221 was used for immunofluorescence assays. For assays requiring a higher number of cells, such as viability analysis by flow cytometry (Annexin V/PI) and protein analysis by Western blot, the 2-chamber 844 chip—characterized by a larger culture area and higher cell yield—was used in both microfluidic systems. For each assay, the same chip type was consistently used for treatments with both 10 µM and 20 µM Palbociclib, enabling direct comparison between lower and higher drug concentrations. Experimental conditions (Flux+, Flux−, Static+, and Static−) were distributed among the chambers as illustrated in Figure 1H, ensuring direct comparability across experimental groups and platforms. For practical reasons, assays using 20 µM Palbociclib were performed with the 2-chamber 844 chips on the incubator-based platform, whereas assays with 10 µM were carried out on the bench-top system.

To enable chip assembly and connection to the syringe-pump systems, standard components from the company ChipShop were used. Fluidic interfaces were established using PTFE tubing (ID 0.5 mm, OD 1 mm; cat. 10000032) and silicone microfluidic tubing (ID 0.5 mm, OD 2.5 mm; cat. 10000033). Chip ports were connected with single male Miniluer connectors (cat. 10000116), individual Miniluer plugs (cat. 10000054), and four-plug Miniluer strips (cat. 10000056). Syringes were coupled to the tubing through polypropylene syringe adaptors (cat. 10000360). These components ensured stable and reproducible fluidic connections throughout all experiments, they are showed in

Table 1. List of chips and components used to settle each chip.

Material	Name	Manufacturer	Catalog Number
Chip 844 with 2 chambers	Rhombic chamber chip, Fluidic 844 Zeonor hydrophilized	ChipShop	10001002
Chip 221 with 4 chambers	Rhombic chamber chip, Fluidic 221 Zeonor hydrophilized	ChipShop	10000293
PTFE tubes	PTFE tubes (ID: 0.5 mm, OD: 1 mm, 1 m)	ChipShop	10000032
Silicon Tubes	Silicone tube for microfluidics (ID: 0.5 mm, OD: 2.5 mm, 1 m)	ChipShop	10000033
Connectors	Microfluidic Chip Connector—Single Male Miniluer Connector, opaque, material: TPE	ChipShop	10000116
Independent Plugs	Single Male Miniluer Plugs, opaque, material: TPE	ChipShop	10000054
Plugs in a row	Row of four Male Miniluer Plug, opaque, material: TPE	ChipShop	10000056
Syringe Adaptors	Syringe adapter. Material: PP	ChipShop	10000360

.

2.2. Calculation of Wall Shear Stress (τw) Inside Chambers of Chip 221 and Chip 844

The wall shear stress (τw) was calculated for the rhombic chambers of ChipShop Fluidic 221 and 844 chips, considering an infusion rate of 20 µL/h. All calculations are SI-consistent and follow a parallel-plate (rectangular-shape) approximation for each chip.

2.2.1. Input Parameters

Infusion rate:

Q = 20 µL/h

Dynamic viscosities:

−

DMEM + 10% FBS: μ = 0.94 mPa·s = 0.94 × 10⁻³ Pa·s

−

RPMI-1640 + 10% FBS (mean): μ = 0.958 mPa·s = 0.958 × 10⁻³ Pa·s

Rhombic chamber geometries:

−

Chip 221: width w = 4.5 mm; height h = 600 µm

−

Chip 844: width w = 7.6 mm; height h = 1.5 mm

2.2.2. Governing Equations

Flow-rate conversion:

Q = 20 µL/h = 20 × 10⁻⁹ m³/h = (20 × 10⁻⁹)/3600 m³/s = 5.556 × 10⁻¹² m³/s

Wall shear stress (parallel-plate/rectangular-shape approximation):

τw = (6 μ Q)/(w h²)

Unit conversion:

1 Pa = 10 dyne/cm²

Step-by-step SI-consistent calculations

Chip 221 (w = 4.5 mm = 4.5 × 10⁻³ m, h = 600 µm = 6 × 10⁻⁴ m)

DMEM + 10% FBS (μ = 0.94 × 10⁻³ Pa·s):

τw = (6 × 0.94 × 10⁻³ × 5.556 × 10⁻¹²)/(4.5 × 10⁻³ × (6 × 10⁻⁴)²)

τw = 1.934 × 10⁻⁵ Pa = 1.934 × 10⁻⁴ dyne/cm²

RPMI-1640 + 10% FBS (μ = 0.958 × 10⁻³ Pa·s):

τw = 1.971 × 10⁻⁵ Pa = 1.971 × 10⁻⁴ dyne/cm²

Chip 844 (w = 7.6 mm = 7.6 × 10⁻³ m, h = 1.5 mm = 1.5 × 10⁻³ m)

DMEM + 10% FBS (μ = 0.94 × 10⁻³ Pa·s):

τw = (6 × 0.94 × 10⁻³ × 5.556 × 10⁻¹²)/(7.6 × 10⁻³ × (1.5 × 10⁻³)²)

τw = 1.832 × 10⁻⁶ Pa = 1.832 × 10⁻⁵ dyne/cm²

RPMI-1640 + 10% FBS (μ = 0.958 × 10⁻³ Pa·s):

τw = 1.867 × 10⁻⁶ Pa = 1.867 × 10⁻⁵ dyne/cm²

2.2.3. Interpretation

These calculations demonstrate that, for an infusion rate of 20 µL/h, the wall shear stress in the rhombic chambers is on the order of 10⁻⁵ Pa for Chip 221 and 10⁻⁶ Pa for Chip 844, as shown in

Table 2. Explanation of each variable (SI units).

Symbol of the Variable	Name of the Variable	SI Unit
τw -	Wall shear stress Pa	(N/m2)
µ-	Dynamic viscosity of the medium	(Pa·s)
Q-	Volumetric flow rate	(m3/s)
w-	Width of the rhombic chamber	(m)
h-	Height (depth) of the rhombic chamber	(m)

. The larger chamber height of Chip 844 leads to significantly lower shear stress due to the quadratic dependence on h in the denominator. The Wall shear stress (τw) calculations for chips 844 and 221 infused with mediums DMEM or RPMI-1640 are showed in

Table 3. Wall shear stress (τw) calculations for chips 844 and 221 infused with mediums DMEM or RPMI-1640.

Chip	Width (mm)	Height (mm)	Medium	Viscosity (Pa·s)	τw (Pa)	τw (dyne/cm2)
221	4.5	0.6	DMEM + 10% FBS	0.94 × 10−3	1.934 × 10−5	1.934 × 10−4
221	4.5	0.6	RPMI-1640 + 10% FBS	0.958 × 10−3	1.971 × 10−5	1.971 × 10−4
844	7.6	1.5	DMEM + 10% FBS	0.94 × 10−3	1.832 × 10−6	1.832 × 10−5
844	7.6	1.5	RPMI-1640 + 10% FBS	0.958 × 10−3	1.867 × 10−6	1.867 × 10−5

.

2.3. Cell Culture Conditions and Reagents

MCF-7 and T47D cells were purchased from Rio de Janeiro Cell Bank- Banco de Células do Rio de Janeiro BCRJ (cat. 0162 and cat. 0323, respectively). MCF-7 cells were maintained in D-MEM/F-12 (cat. 12400-024, Gibco, Waltham, MA, USA) in t75 flasks in a confluence of 7.5 × 10⁵ cells in 10 mL medium (7.5 × 10⁴ cells/mL), and T47D cells were maintained in RPMI 1640 (cat. 31800-022, Gibco) in t75 flasks in a confluence of 7.5 × 10⁵. Both media were supplemented with 10% fetal bovine serum (FBS, cat. 12657-029, Gibco), 1% L-Glutamine (cat. 21051-024, Sigma, St. Louis, MO, USA), and 1% Penicillin/Streptomycin (P/S, cat. 15140-122, Gibco). Both cells were used until passage 30. For all assays, the following groups were evaluated: Palbociclib treated cells submitted to the fluidic/infusion rate of 20 µL/h of the microfluidic platform/syringe pump (Treated + F), non-treated cells with same influx rate (NT − F), treated cells under static condition (Treated + S) and non-treated cells under static condition (NT − S). For treated cells, the antitumor drug Palbociclib Isethionate (1000 µM stock-solution, cat. S1579, Selleck Chemicals LLC, Houston, TX, USA) was diluted in cell medium (D-MEM for MCF-7 cells or RPMI for T47D cells) to achieve 10 or 20 μM concentrations. Palbociclib was chosen for the standardization of the microfluidic platforms because it is a highly selective inhibitor of CDK4/6 kinases and a well-established, reliable treatment [38,39].

2.4. Ki67 Proliferation Assay

To evaluate if the treatment with Palbociclib was properly affecting cell proliferation in both cell lineages, we performed a Ki67 proliferation assay as described previously [40]. MCF-7 and T47D proliferating cells were identified with anti-Ki67 antibody (1:300, cat. MA-514520, Invitrogen, Carlsbad, CA, USA) in association with DAPI staining. The MCF-7 and T47D cells were seeded in a microfluidic chip 221 (ChipShop, Jena, Germany) at a concentration of 1 × 10⁵ cells per chamber and in a chip 844 (ChipShop, Jena, Germany) at a concentration of 2 × 10⁵ cells per chamber. The following day, cells were treated or not with 10 μM or 20 μM of the antitumor drug Palbociclib for 48 h. Cells seeded in the 221 chip (1 × 10⁵ cells/chamber) received the lowest concentration of Palbociclib (10 μM), and cells in the 844 chip (2 × 10⁵ cells/chamber) received a higher concentration of the drug (20 μM). After 48 h of treatment, cells were washed twice with PBS, fixed inside the chips with 4% PFA for 20 min at room temperature (RT), and permeabilized and blocked with PBS-BSA1%-Triton 0.5% for 30 min at room temperature. Both cells were then incubated with anti-Ki67 overnight under shaking, followed by anti-rabbit Alexa Fluor^TM 488 donkey anti-rabbit IgG secondary antibody (1:5000, cat. A21206, Invitrogen) staining for 1 h under shaking at room temperature (RT). Next, cells were washed and incubated with 4′,6-diamidino-2-phenylindole (DAPI) for 20 min at RT. These samples were analyzed in an Operetta CLS^TM High Content Analysis System (Perkin Elmer, Waltham, MA, USA) device. Data were analyzed using the Harmony^® 4.8 High Content Imaging and Analysis Software, and the statistical analysis/graphical representations were created with GraphPad Prism 9.0.0 software. Images were also acquired using a Leica DMI6000B fluorescence microscope (Leica Microsystems GmbH, Wetzlar, Germany) with LAS AF software 3.0. at the Confocal and Electron Microscopy Platform (RPT07C) of Carlos Chagas Institute, Fiocruz/PR, Brazil. Statistical analysis was performed using Graphpad Prism software version 9.0.0. Data normality was first assessed using the Kolmogorov–Smirnov test. When the data did not follow a normal distribution, comparisons between treated and untreated groups were performed using the Kruskal–Wallis test. Results are presented as mean ± standard deviation (SD). Statistical significance was defined as p < 0.05.

2.5. Annexin V Cell Viability Assay by Flow Cytometry

To assess the effect of Palbociclib on cells cultured within the microfluidic BoC model, apoptosis was quantified using the Dead Cell Apoptosis Kit (cat. V13241, Invitrogen), which enables discrimination of viable, early apoptotic, and late apoptotic/necrotic cells based on phosphatidylserine externalization and membrane permeability. Cells were seeded into two-chamber 844 microfluidic chips (ChipShop) at densities of 2 × 10⁵ cells/chamber for the 10 µM Palbociclib condition and 4 × 10⁵ cells/chamber for the 20 µM Palbociclib condition. After 24 h, chips were connected to syringes containing either culture medium alone (untreated control) or medium supplemented with Palbociclib (treated groups). Syringes and chips were then mounted onto the microfluidic syringe pump, and cells were exposed to continuous Palbociclib treatment for 48 h. Following treatment, cells were enzymatically detached from the chips using TrypLE™ (Gibco, 12604-013) for 5 min at 37 °C, after which the reaction was quenched with medium containing 10% FBS. Cell suspensions were transferred to microtubes, centrifuged at 3000× g for 5 min, and the supernatant was carefully removed. Staining was performed following the kit instructions. Briefly, pellets were resuspended in 100 µL of 1× Annexin-Binding Buffer, followed by addition of 5 µL Alexa Fluor™ 488 Annexin V and 1 µL propidium iodide (100 µg/mL). Samples were gently mixed and incubated for 15 min at room temperature in the dark. After incubation, 400 µL of 1× Annexin-Binding Buffer was added, and samples were kept on ice until acquisition. Single color compensation and experimental positive controls were included to ensure accurate gating. For these controls, cells were lysed at 60 °C for 30 min and subsequently stained with 5 µL Annexin V or propidium iodide for 15 min. Of all samples and controls, 10,000 events were analyzed immediately after staining using a CytoFLEX LX N0-V5-B3-Y5-R3-I0 flow cytometer (Beckman Coulter, Brea, California, USA) equipped with 488 nm and 594 nm lasers at the Flow Cytometry Platform (RPT 08L) of Carlos Chagas Institute, FIOCRUZ/PR. Data were processed using FlowJo v10.10. For the viability assay, comparisons between treated and untreated groups were initially performed using a parametric Student’s t-test. When data did not meet normality assumptions as assessed by the Shapiro–Wilk test, the non-parametric Mann–Whitney U test was applied to all samples. Analyses were conducted using a 95% confidence level, and data are presented as mean ± standard deviation (SD). Differences were considered statistically significant at p < 0.05. Statistical analyses were performed using GraphPad Prism (version 9.0.0).

2.6. PARP1 Resistance Marker Processing Analysis by Western Blot (WB)

PARP1 processing in MCF-7 and T47D resistance markers was investigated using SDS-PAGE and Western blot, as previously published [40]. Cells were trypsinized from the chips and then lysed in RIPA Buffer (50 mM Tris-HCl; 150 mM NaCl; 0.5% SDS; 0.5% Triton X-100; pH 7.2) for 20 min on ice. Sample buffer containing 4X bromophenol blue and β-mercaptoethanol (Laemmli buffer) was added to samples (30 μg of whole cell lysate, WCL), which were then denatured at 95 °C for 5 min. Cells’ protein extracts were loaded onto a 13% polyacrylamide gel and separated at 20–30 mA in a tank containing running buffer (25 mM Tris, 192 mM glycine, and 0.1% SDS). The transfer of proteins from the gel to the nitrocellulose membrane (GE Healthcare, Hertfordshire, UK) was carried out at 23 volts for 1 h in a Trans-Blot^® SD Semi-Dry Transfer Cell with transfer buffer (25 mM Tris; 192 mM glycine; 20% methanol (v/v)) chilled in the refrigerator. Proteins on the membrane were visualized by staining with 0.2% Ponceau, and the membranes were subsequently blocked with 5% TBST milk solution (200 mM Tris-HCl (pH 7.5); 1.5 M NaCl; 0.05% of Tween 20; and 5% non-fat dry milk) for 1 h at room temperature. PARP1, used as a resistance marker, was detected with an anti-PARP1 antibody (1:1000; cat. AB227244, Abcam, Cambridge, UK), while total protein loading control was assessed using an anti-GAPDH antibody (1:1000; cat. sc-47724, Santa Cruz Biotechnology, Dallas, TX, USA).

Antibodies were diluted in 0.1% TBST-BSA solution. Primary antibody incubation was performed overnight at 4 °C under gentle agitation. Membranes were then washed three times with 0.05% TBST for 5 min each. The secondary antibody, Alexa Fluor™ 488 donkey anti-rabbit IgG (1:5000; cat. A21206, Invitrogen), was incubated for 1 h at room temperature under shaking. Afterwards, membranes were washed three times with 0.05% TBST for 5 min. Bands were detected using the iBright™ imaging system (Thermo Fisher Scientific, Waltham, MA, USA). Western blot assays were performed 3 times (n = 3) for each cell line, and the most representative picture was selected.

2.7. PARP1 Resistance Marker Detection by Immunofluorescence

To evaluate if treatment with Palbociclib was properly affecting cell resistance, an immunofluorescence assay was performed for the detection of resistance marker PARP1. PARP1 was detected in MCF-7 and T47D with anti-PARP1 antibody (1:200, cat. AB227244, Abcam) and Alexa 488 secondary anti-rabbit in association with DAPI staining. The MCF-7 and T47D cells were seeded in a microfluidic chip 221 (ChipShop, Jena, Germany) at a concentration of 2 × 10⁵ cells per chamber and in a chip 844 (ChipShop, Jena, Germany) at a concentration of 4 × 10⁵ cells per chamber. The following day, cells were treated or not with 10 μM or 20 μM of the antitumor drug Palbociclib for 48 h. Cells seeded in the 221 chip (2 × 10⁵ cells/chamber) received the lowest concentration of Palbociclib (10 μM), and cells in the 844 chip (4 × 10⁵ cells/chamber) received the highest concentration of the drug (20 μM). After 48 h of treatment, cells were washed with PBS, fixed inside the chips with 4% PFA for 20 min at room temperature (RT), and permeabilized and blocked with PBS-BSA1%-Triton 0.5% for 30 min at room temperature. Both cells were then incubated with anti-PARP1 overnight under shaking and then reacted with anti-rabbit Alexa Fluor^TM 488 donkey anti-rabbit IgG secondary antibody (1:5000, cat. A21206, Invitrogen) for 1 h under agitation at RT. Next, cells were washed and incubated with DAPI for 20 min at RT. Images were acquired with a fluorescence Leica DMI6000B microscope and LAS AF software 3.0 at the Confocal and Electron Microscopy Platform (RPT07C) of Carlos Chagas Institute, FIOCRUZ/PR.

2.8. Beta-Tubulin Cytoskeleton Analysis by Immunofluorescence

In order to assess whether treatment with Palbociclib was able to disarrange cytoskeleton molecules in both lineages, we performed an immunofluorescence assay staining cytoskeleton marker β-tubulin. The MCF-7 and T47D cells were seeded in a microfluidic chip 221 (ChipShop, Jena, Germany) at a concentration of 1 × 10⁵ cells per chamber and in a chip 844 (ChipShop, Jena, Germany) at a concentration of 2 × 10⁵ cells per chamber. On the following day, cells were treated with Palbociclib (10 or 20 μM) or left untreated for 48 h. Cells seeded in the 221 chip (1 × 10⁵ cells/chamber) received 10 μM of Palbociclib, and cells in the 844 chip (2 × 10⁵ cells/chamber) received 20 μM of the drug. After 48 h, cells were fixed inside the chips with 4% PFA for 20 min at RT, permeabilized, and blocked with PBS-BSA1%-TRITON 0.5% for 30 min at room temperature. MCF-7 and T47D cytoskeletons were incubated with anti-β-tubulin antibody (cat. PA5-16863, Invitrogen) overnight under shaking, followed by anti-rabbit Alexa Fluor^TM 488 donkey anti-rabbit IgG secondary antibody (cat. A21206, Invitrogen) staining for 1 h under shaking at RT. Next, cells were washed and incubated with DAPI for 20 min at RT. Images were acquired using an Operetta CLS^TM High Content Analysis System (Perkin Elmer, Waltham, MA, USA) device and the Harmony^® 4.8 High Content Imaging and Analysis Software. Additional images were acquired with fluorescence microscope Leica DMI6000B and LAS AF software 3.0 at the Confocal and Electron Microscopy Platform (RPT07C) of Carlos Chagas Institute, FIOCRUZ/PR.

2.9. Statistical Analysis

For Ki67 proliferation assays, data distribution was tested for normality using the Kolmogorov–Smirnov and Shapiro–Wilk parametric tests. However, some samples did not show normal distribution; therefore, all samples were submitted to a non-parametric test. Kruskal–Wallis comparison between treated and non-treated groups (mean ± SD) was conducted using a 95% confidence interval. Differences were considered statistically significant at p < 0.05 (*).

For Annexin V cell viability assays, data distribution was assessed using the Shapiro–Wilk normality test. As the data did not follow a normal distribution, comparisons between treated and non-treated groups were performed using the non-parametric Mann–Whitney test. Data were presented as mean ± standard deviation (mean ± SD), with a 95% confidence interval. Differences were considered statistically significant at p < 0.05 (*). Statistical analysis was performed using Graphpad Prism software version 9.0.0.

3. Results

3.1. Palbociclib Inhibited Proliferation of MCF-7 and T47D Cells

To evaluate whether the new breast-on-a-chip device could be used for proliferation assays and if Palbociclib could interfere with cell proliferation when administered in this system, a Ki67 assay was performed. A reduction in cell proliferation was observed in treated MCF-7 cells under fluidic conditions (Figure 2C,E), especially under treatment with the 20 µM Palbociclib concentration (Figure 2B,E,H) when compared to the 10 μM (Figure 2A,C,G). Under static conditions (S), which serve as a reference condition, proliferation was also further reduced in cells treated with 20 μM Palbociclib (Figure 2F,H) than in those treated with 10 μM (Figure 2D,G). No significant differences were observed between treated groups under microfluidic versus static conditions.

Proliferation of T47D cells was also evaluated through the Ki67 assay. A significantly reduced proliferation was observed in T47D cells treated with a higher concentration of Palbociclib (20 µM) (Figure 3B,E,H) compared with those treated with 10 µM (Figure 3A,C,G) under both fluidic (Figure 3C,E) and static conditions (Figure 3D,F). No significant differences in proliferation were detected between treated groups under microfluidic versus static conditions.

3.2. Palbociclib Indirectly Reduced Viability of MCF-7 and T47D Cells

The Annexin V flow cytometry was used to assess whether treatment with Palbociclib under microfluidic conditions altered cell viability, apoptosis, or necrosis. No significant differences were observed in early apoptosis, cell death, or necrosis in MCF-7 cells treated with 10 μM Palbociclib (Figure 4A–D,M–P) compared with the non-treated control group (Figure 4I–L). In contrast, treatment with 20 μM Palbociclib (Figure 4E–H,Q–T) induced a significant increase in late apoptotic (dead) cells relative to the non-treated group, although no differences were detected in early apoptosis or necrosis.

In accordance with the findings for MCF-7 cells, T47D cells showed no significant differences in early apoptosis, cell death, or necrosis between those treated with the lower concentration of Palbociclib (10 μM) (Figure 5A–D,M–P) and the non-treated control group (Figure 5I–L). Similarly, treatment with 20 μM Palbociclib (Figure 5E–H,Q–T) induced a significant increase in late apoptotic (dead) cells compared with the non-treated group, while no significant differences were observed in early apoptosis or necrosis.

3.3. Palbociclib Indirectly Affects the Processing of Resistance Marker PARP1 in MCF-7 and T47D Cells

Western blot and immunofluorescence assays were carried out to evaluate whether treatment with Palbociclib in the breast-on-a-chip (BoC) system would alter the processing of the resistance marker PARP1. A total of four independent assays showed the same pattern of PARP1 processing in the resistant lineage MCF-7 after treatment, with prevalence of 116 kDa full-length non-cleaved PARP1 (Figure 6A and Figure S1, ncPARP1) associated with cell survival even after treatment with a higher concentration of Palbociclib (20 μM) (Figure 6A, 20 µM; Figure 6D). This suggests that exposure to higher drug concentrations could inhibit PARP1 cleavage.

In the drug-sensitive T47D cells, both non-cleaved 116 kDa-PARP1 (ncPARP1) and cleaved 89 kDa-PARP1 (cPARP1) (Figure 6B and Figure S2) were detected after treatment with both concentrations of Palbociclib (Figure 6B, 10 µM and 20 µM; Figure 6E,F). However, the cleaved form of 89 kDa (cPARP1) associated with cell death and apoptosis was more predominant in this cell lineage compared to MCF-7 cells, suggesting an increased drug sensitivity of T47D cells to Palbociclib. Through band densitometry analysis, no significant expression of non-cleaved PARP1 (ncPARP1) was observed in treated and non-treated groups of MCF-7 (Supplementary Figure S1E,F) and T47D cells (Supplementary Figure S2E,F) treated with 10 or 20 µM Palbociclib.

3.4. Palbociclib Induced the Cytoskeleton Disarrangement in MCF-7 and T47D Cells

MCF-7 and T47D cells were treated with Palbociclib to examine its impact on cytoskeletal organization and thus evaluate whether the BoC model was suitable for assessing drug effects. An immunofluorescence assay targeting the cytoskeleton marker β-tubulin was performed. Healthy MCF-7 and T47D cells typically form dome-like structures with cuboid morphology. In the BoC system, MCF-7 cells treated with both 10 and 20 μM Palbociclib showed evident morphological alterations, including loss of cuboid shape, disruption of dome-like structures, and disorganization of the β-tubulin cytoskeleton (Figure 7). These changes were more pronounced in cells treated with 20 μM Palbociclib, as evidenced by a stronger disruption of β-tubulin architecture (Figure 7B, upper panel, Figure 7E,F) compared with non-treated controls and the 10 μM treatment group under static conditions (Figure 7A,C,D), which retained more intact dome-like structures.

Disorganized cytoskeleton, disrupted dome-like structures, and a less cuboid morphology were also observed in T47D cells treated with the higher concentration of Palbociclib (20 μM) (Figure 8B, upper panel; Figure 8E,F) when compared with non-treated cells and those treated with the lower concentration (10 μM) (Figure 8A,C,D). Thus, in both MCF-7 and T47D cells, Palbociclib treatment induced disruption of cell-to-cell interactions, loss of structural cohesion, and disarrangement of dome-like architectures. In both cell lines, these effects were more pronounced at 20 μM of Palbociclib.

4. Discussion

In this study, we validated a new 2D BoC model using two luminal breast cancer cell lines, MCF-7 and T47D, as a controlled platform for reproducible evaluation of drug responses under continuous perfusion. By using Palbociclib as a reference compound with established biological effects, we demonstrated that the platform consistently captures proliferation, viability, cytoskeletal alterations, and resistance-associated marker dynamics under defined microfluidic conditions. The new 2D BoC system supported long-term culture under flow, preserved cell morphology and viability, and allowed the analysis of multiple drug response readouts, including proliferation, apoptosis, cytoskeletal organization, and processing of the resistance-associated marker PARP1. These features indicate that the model is suitable for studying drug effects in a controlled microenvironment that is closer to in vitro tissue-like conditions than conventional static culture. The shear stress in capillary blood vessels is greatly variable depending on vessel diameter, hematocrit, and local flow conditions. In vivo measurements and modeling studies report capillary wall shear stresses ranging from 10⁻⁶ to 10⁻⁴ Pa, particularly in plasma-dominated or tumor microvasculature [41,42,43]. Considering the infusion rate of 20 µL/h, the wall shear stress in the rhombic chambers was on the order of 10⁻⁵ Pa for Chip 221 and 10⁻⁶ Pa for Chip 844, values that fall within the lower physiological range of capillary shear stress and are especially representative of breast tumor microvascular conditions characterized by slow and intermittent perfusion [44]. Moreover, in our closed microfluidic system, no meaningful alterations in pH or phenol red media color were observed over 48 h in either DMEM or RPMI cultures of MCF-7 and T47D cells (Supplementary Figures S3 and S4). Direct pH measurements confirmed stable conditions, consistent with minimal gas exchange in the sealed devices, which preserves the bicarbonate–CO₂ equilibrium and maintains physiologically relevant pH despite the absence of external CO₂ supplementation.

Although different chip geometries (221 vs. 844) and pump setups were employed, all experiments were conducted under identical infusion rates (20 µL/h), continuous perfusion, and controlled temperature, resulting in low and comparable capillary-level shear stress (on the order of 10⁻⁵ Pa for Chip 221 and 10⁻⁶ Pa for Chip 844). Under these controlled dynamic conditions, consistent biological trends in proliferation, viability/apoptosis, and cytoskeletal alterations were observed in both MCF-7 and T47D cells across experimental configurations. While this does not exclude subtle effects of chamber geometry or media volume on drug transport, it supports the interpretation that the main conclusions of this study are not driven by major platform-related effects. Accordingly, this work focuses on validating the robustness of the dynamic 2D breast-on-a-chip model rather than claiming platform independence or quantitative equivalence between microfluidic systems.

In our BoC model, Palbociclib treatment reduced proliferation in both MCF-7 and T47D cells, with a stronger effect at the concentration of 20 µM. Palbociclib is a CDK4/6 inhibitor commonly used in the treatment of luminal A–like (ER⁺/HER2⁻) breast cancer, where it reduces cell proliferation by inducing G1 cell-cycle arrest. Our findings corroborate many other works using Palbociclib to treat breast cancer cells [25,26,27] specifically luminal T47D cells [38], triple negative breast cancer cells (TBNCs), despite these being less sensitive to Palbociclib than luminal cells [45] and different cancer cells [18,33,34] or other similar CDK4/6 inhibitors such as Dinaciclib [21]. Harada and colleagues (2018) showed that in T47D cells, Palbociclib enhances SMAD2 binding to the genome by inhibiting CDK4/6-mediated linker phosphorylation of the SMAD2 protein and consequently inhibiting cell proliferation [38]. Although some authors used cell culture models different from OoCs, they all demonstrate that Palbociclib reduces cell proliferation. Our 2D BoC model reproduced the increased efficacy of Palbociclib when administered in a microfluidic chip, in accordance with previous works showing increased efficacy of different drugs administered in 3D breast-on-a-chip microfluidic models [19]. These authors showed the increased efficacy of different drugs in lower concentrations (e.g., Paclitaxel and Olaparib) administered in a 3D breast-on-a-chip model using different breast cancer cells (e.g., triple negative breast lineages, MDA-MB-453, MDA-MB-231, and HCC1937), highlighting the importance of a microfluidic BoC model [19].

Our findings regarding Palbociclib’s effect on cell proliferation are in accordance with the findings of Testa et al. 2025, who used a different breast-on-a-chip/tumor-on-chip model with primary ductal carcinoma breast cancer cells, HCC1937 cells treated with Dinaciclib (a cyclin-dependent kinase (CDK) inhibitor similar to Palbociclib), showing a significantly reduced proliferation of cancer cells [21]. Likewise, for our BoC model, Testa et al. (2025) showed that perfusion in the tumor-on-chip model without an antitumor drug was able to preserve the growth ability and morphology of cancer cells [21]. Additionally, these findings corroborate our results showing that treatment of MCF-7 and T47D cells with Palbociclib in a microfluidic BoC model reduced cell proliferation, increased cell death, and induced cytoskeleton disarrangement.

Regarding the viability of MCF-7 and T47D cells after Palbociclib treatment in the microfluidic BoC model, annexin V assays demonstrated that the 20 µM dose increased the proportion of late apoptotic or dead cells, particularly in T47D cells, revealing a differential sensitivity between the two cell lines. These findings are consistent with the expected pharmacological action of a CDK4/6 inhibitor and demonstrate that our BoC model is able to detect dose-dependent and lineage-specific responses to Palbociclib. Additionally, our annexin V assay findings corroborate previous reports from other breast-on-a-chip models using MCF-7 cells treated with a different drug, doxorubicin, regarding increased cytotoxicity, reduced cell viability, and increased cell death evaluated by LIVE/DEAD and lactate dehydrogenase (LDH) assays [46]. Our data related to reduced cell viability after treatment with Palbociclib in our BoC model are in accordance with the findings of other work using a different layered breast-cancer-on-a-chip system to study cytotoxicity of doxorubicin (for breast cancer cells), which showed that microfluidic flow enhanced the effect of this drug in ductal breast cancer MCF-7 cells, leading to reduced cell viability. [22]. Similarly to previous studies demonstrating that organ-on-a-chip models can be used to investigate drug penetration, tumor–stromal interactions, and anticancer drug efficacy [22], our results indicate that the present breast-on-a-chip model provides a promising platform for antitumor drug assessment using a simplified experimental setup. We also observed reduced viability of MCF-7 cells treated with Palbociclib in the breast-on-a-chip model, consistent with previous reports showing that microfluidic systems enhance drug efficacy in MCF-7 cells. [22].

Cytoskeletal staining further supported these observations. In both cell lines, Palbociclib induced disorganized beta tubulin networks, loss of cuboid morphology, and disruption of dome-like structures, changes that were more pronounced at 20 µM. These structural alterations suggest impaired cell-to-cell interactions and reduced epithelial integrity, adding a morphological dimension to the functional readouts of drug response in the BoC system. Moreover, β-tubulin cytoskeleton conformation disarrangement, dome-like structures disarrangement, and less cuboid morphology were more prominent in MCF-7 treated with 20 μM Palbociclib under fluidic conditions compared to the non-treated group and group treated with 10 μM of the drug and in static conditions. In both lineages, the treatment with a higher dose of Palbociclib induced more disruption of cell-to-cell interactions and detachment between cells. This altered morphology following treatment with Palbociclib is in accordance with previous findings using patient-derived glioma stem cell-enriched cell lines, in which Palbociclib rapidly and effectively inhibited proliferation through CDK6 inhibition and induced a senescent-like quiescent phenotype characterized by flattened morphology, cell cycle arrest, increased β-galactosidase activity, and induction of other senescent-associated markers [47].

To explore the association between PARP1 and cell fate in MCF-7 and T47D cells, we analyzed PARP1 processing and observed the presence of both full-length and cleaved PARP1 in both cell lines. It is known that PARP1 is cleaved into two fragments: the cleaved 89 kDa PARP1 fragment generated by caspase-3 that promotes DNA fragmentation in caspase-mediated apoptosis and the 24 kDa fragments that are associated with DNA breaks [37]. Therapies like Palbociclib in MCF-7 and T47D cells generate a functional BRCAness phenotype in BRCA1/2-wt cells by downregulating key DNA repair proteins that leads to inhibition of global transcription and cell death in a PARP1-dependent manner [48]. This agrees with our findings showing predominance of cleaved PARP1 and cell death in breast cancer cells treated with higher doses of Palbociclib (20 µM). Analysis of PARP1 processing revealed additional differences between the two lines. In MCF 7 cells, the predominant band corresponded to the full-length non-cleaved 116 kDa form (ncPARP1), which is associated with cell survival and resistance. In T47D cells, both the 116 kDa full-length protein and the 89 kDa cleaved fragment (cPARP1) were detected, and Palbociclib treatment increased the relative abundance of the cleaved form. Since the 89 kDa fragment is linked to apoptotic processes, this pattern is compatible with the higher sensitivity of T47D cells seen in the proliferation and Annexin V assays. Although PARP1 was quantified by a low-sensitivity method (Western blot band densitometry) and no significant PARP1 expression was observed between treated and non-treated groups in both cell lines (Supplementary Figures S1E,F and S2E,F), the qualitative changes observed illustrate that the BoC model can capture processing variations in resistance-related markers. Hence, we suggest that Palbociclib indirectly induces apoptosis in MCF-7 and T47D breast cancer cells and PARP1 caspase-mediated fragmentation in T47D cells, but whether the cell DNA fragmentation is AIF-mediated remains to be further investigated.

We observed differential processing patterns of PARP1, with more predominant 116 kDa full-length PARP1 in the resistant MCF-7 lineage, consistent with previous studies showing that non-cleaved PARP1 associated with cell survival is upregulated in BRCA1-mutated resistant tumors [49]. We observed both non-cleaved 116 kDa-PARP1 and cleaved 89 kDa-PARP1 in T47D cells. This corroborates previous findings showing both forms in the majority of BRCA1-associated breast cancers [50]. In the T47D sensitive cell lineage, we observed low expression of full-length PARP1, consistent with findings seen for cancers with increased sensitivity, like BRCA1-protein-deficient ovarian cancer [51]. In sensitive luminal T47D cells, an increased expression of cleaved PARP1 was observed, indicating increased cell death. Hence, our findings are in accordance with previous findings that demonstrate that the expression levels of the PARP-1- and estrogen-co-regulated gene set are enriched in the luminal subtype of breast cancer, and high PARP-1 expression in ER+ cases is related to reduced survival [52]. In summary, although p53 mutations present in T47D cells are frequently associated with therapy resistance, p53 signaling was not directly interrogated in this study; therefore, Palbociclib response in the breast-on-a-chip model was better captured by functional biomarkers such as PARP1 processing rather than p53 status alone.

Although Palbociclib responses in MCF-7 and T47D cells are well characterized in static 2D and 3D cultures, few studies have evaluated this drug under microfluidic conditions. While no significant proliferation differences were observed between static and microfluidic cultures, the value of the breast-on-a-chip model lies in its physiological relevance rather than altered proliferation outcomes. Controlled continuous perfusion enables dynamic drug delivery, nutrient exchange, waste removal, and capillary-level shear stress that better recapitulate in vivo microvascular conditions. Beyond Palbociclib testing, this study validates the platform’s robustness and adaptability for future anticancer drug screening, including potential applications with patient-derived biopsy samples.

The scarcity of detailed mechanistic data from 3D tumor-on-a-chip models suggests that simplified microfluidic systems can provide sufficient and relevant information for antitumor drug evaluation while serving as a foundation for future development of more complex 3D platforms. Thus, Palbociclib testing in organ-on-a-chip and breast-on-a-chip models warrants further investigation.

An important finding of this study is that comparable biological responses were obtained using two different microfluidic setups operated under identical infusion rate and temperature conditions. Despite differences in pumping and temperature-control platforms, cell behavior and drug responses were consistent, indicating that the observed effects are intrinsic to the chip-based culture rather than equipment-dependent. This demonstrates the robustness and reproducibility of the BoC model and supports its implementation across different laboratory infrastructures.

Overall, the results show that the breast-on-a-chip model described here is a useful platform for evaluating the effects of Palbociclib on breast cancer cells, integrating information from viability, proliferation, morphology, cytoskeletal organization, and PARP1 processing. As a next step, the model can be expanded by testing other drugs or, for example, incorporating three-dimensional cultures such as spheroids or organoids, by including additional breast cancer subtypes, and introducing co-cultures with stromal and immune cells to better represent the tumor microenvironment. The use of patient-derived cells and drug combinations, such as CDK4/6 and PARP inhibitors, can further increase the translational relevance of this system. Despite these future developments, the current study already establishes a solid and reproducible BoC platform that other groups can adapt for systematic testing of antitumor drugs under standardized microfluidic conditions.

Some limitations of this study should be acknowledged. (1) Two-dimensional model—the study used a 2D microfluidic culture system without incorporating 3D models (e.g., spheroids or organoids), which better reproduce breast tissue architecture, cell–cell interactions, and in vivo–like drug absorption. (2) No comparison with static cultures—a direct comparison with conventional static 2D plate cultures was not performed, limiting the evaluation of the added value of the microfluidic platform. (3) Restricted subtype representation—only two luminal A–like cell lines (MCF-7 and T47D) were analyzed, limiting extrapolation to other subtypes, including luminal B and more aggressive phenotypes. (4) Absence of negative controls—triple-negative breast cancer cell lines (e.g., MDA-MB-231, MDA-MB-436) and non-tumorigenic epithelial cells (e.g., MCF10A) were not included, restricting broader biological comparisons. (5) No patient-derived samples—primary patient-derived breast cancer cells were not incorporated, limiting assessment of inter-patient heterogeneity and translational relevance. (6) Limited therapeutic evaluation—only a small number of anticancer agents were tested, with Palbociclib as the reference compound; combination therapies and alternative drug classes were not explored. (7) Lack of tumor microenvironment components—the system did not include stromal, fibroblast, endothelial, or immune co-cultures, limiting modeling of tumor–microenvironment interactions. (8) No multi-organ pharmacokinetic assessment—multi-organ-on-a-chip modules (e.g., liver, kidney, intestinal tissues) were not incorporated, precluding evaluation of systemic drug metabolism and pharmacokinetics. (9) Different chip geometries—distinct chip formats (221 and 844) were used for different drug concentrations due to operational constraints; although flow rate, temperature, and shear stress were standardized, formal cross-platform validation was not conducted, and subtle geometrical or volumetric effects on drug transport cannot be excluded. (10) Potential inter-platform variability—experiments were not performed within a single multi-chip incubation platform, possibly introducing subtle variability despite standardized conditions. (11) Limited quantitative molecular validation—although qualitative PARP1 processing patterns were reproducible, a more sensible quantitative validation of PARP1 expression by qPCR was not performed, limiting mechanistic interpretation of resistance-associated marker dynamics and molecular responses to Palbociclib.

Finally, the study focused on relatively short-term exposures and endpoint analyses, and therefore does not capture long-term adaptive responses, acquired resistance mechanisms, or clonal selection that may arise during prolonged treatment. Future studies incorporating 3D culture systems, multi-cellular co-cultures, patient-derived models, expanded drug panels, and integrated multi-organ platforms will be essential to address these limitations and to further enhance the predictive power of breast-on-a-chip microfluidic systems for translational drug testing.

Other perspectives of this work are: the quantification of PARP1 expression by qPCR; the analysis of treated cells’ and their secretome’s molecular content, as well as the creation of 3D spheroid models for drug efficacy assessment. Overall, despite the above-mentioned limitations, this new microfluidic platform allows not only the treatment of cells to evaluate cell morphology and proliferation, but also to research more complex mechanisms related to cancer, such as cell migration, cell viability, and the expression of resistance markers. Lastly, the breast-on-a-chip model favors the treatment combination between the antitumor drug and other drugs/molecular inhibitors of the proliferation signaling pathways, and drug synergy could also be investigated.

5. Conclusions

In this work, we validated a BoC microfluidic model as a controlled platform for drug-response assessment, using Palbociclib as a reference compound in two luminal breast cancer cell lines (MCF-7 and T47D), which display different levels of drug sensitivity. The model enabled the simultaneous evaluation of morphology, viability, proliferation, cytoskeletal organization, and processing of the resistance-associated marker PARP1 under controlled flow conditions.

Palbociclib, at both 10 and 20 µM, reduced cell proliferation and viability and induced cytoskeletal disorganization and loss of dome-like structures in both cell lines. Qualitative analysis of PARP1 suggested lineage-specific responses. Considering PARP1 processing patterns in MCF-7 cells, the full-length fragment (116 kDa) predominantly appeared at the higher drug concentration when compared to the cleaved 89 kDa form, consistent with a more resistant phenotype. In contrast, T47D cells exhibited both full-length (116 kDa) and cleaved (89 kDa) PARP1 fragments, with a predominance of the cleaved form at 20 µM of Palbociclib, in agreement with the greater sensitivity of this cell line and the increase in apoptotic cell death observed by flow cytometry.

Importantly, comparable biological trends in response to Palbociclib were observed when the breast-on-a-chip model was operated using two distinct microfluidic setups under identical flow and temperature parameters. While this does not demonstrate formal equivalence between platforms, it supports the interpretation that the main drug-response profiles were not driven by major differences in pumping configuration. Taken together, these findings support the use of this breast-on-a-chip model as a robust and reproducible in vitro platform for small-scale drug efficacy testing and for the investigation of treatment response and resistance mechanisms in breast cancer cells.

The development of organ-on-a-chip platforms requires a thorough understanding of microfluidic operation and its associated instrumental challenges. The relevance of this study lies in demonstrating that a 2D microfluidic platform can reproducibly integrate multiple drug-response endpoints under physiologically relevant flow, providing a robust and scalable validation framework for future anticancer bioassays.