Deterministic assembly of chromosome ensembles in a programmable membrane trap array

Abstract

Selective spatial isolation and manipulation of single chromosomes and the controlled formation of defined chromosome ensembles in a droplet-based microfluidic system is presented. The multifunctional microfluidic technology employs elastomer valves and membrane displacement traps to support deterministic manipulation of individual droplets. Picoliter droplets are formed in the 2D array of microscale traps by self-discretization of a nanoliter sample plug, with membranes positioned over each trap allowing controllable metering or full release of selected droplets. By combining discretization, optical interrogation, and selective droplet release for sequential delivery to a downstream merging zone, the system enables efficient manipulation of multiple chromosomes into a defined ensemble with single macromolecule resolution. Key design and operational parameters are explored, and co-compartmentalization of three chromosome pairs is demonstrated as a first step toward formation of precisely defined chromosome ensembles for applications in genetic engineering and synthetic biology.

1. Introduction

The manipulation of chromosome ensembles within biological cells offers a powerful route toward the engineering of defined genetic states for targeted organisms. Genetic modification through the addition, removal, or substitution of wild type chromosomes has long been performed in fish and plant breeding to control desired characteristics of the resulting genome [1]. Manipulation of chromosomes within egg cells typically involves shocking a group of cells with a sharp temperature or pressure gradient to induce the addition of a haploid chromosome set from a second polar body [2]. While this approach can modify the chromosome set number by inducing polyploidy, it does not offer a path to engineering an arbitrarily defined ensemble of chromosomes within a given cell. Ongoing advances in the development of synthetic chromosomes have opened the door to a higher level of genetic control. Artificial yeast chromosomes were first demonstrated in 1983 [3], and techniques for the design and bottom-up assembly of fully synthetic eukaryotic chromosomes have been developed in the past decade [4–6]. The construction of a synthetic prokaryotic genome was recently achieved [7], and promising steps toward the realization of synthetic prokaryotic chromosomes have been reported despite the challenges associated with their size and structural complexity [8]. These advances hint at the potential for using chromosome manipulation to engineer new gene circuits, with the potential to vastly expand the tool set available for advancing synthetic biology [8, 9].

As developments in both bottom-up and top-down chromosome engineering advance, there is an emerging need for technologies capable of manipulating and assembling precisely defined chromosome sets. The sorting and isolation of individual chromosomes may be performed by flow cytometry [10, 11], but this approach is cumbersome for the manipulation of large numbers of different macromolecule species, and cannot sequester the final chromosome set within a small volume suitable for cell delivery. As an alternative, various examples of chromosome handling using microfluidic systems have been reported. Microfluidic technologies have been explored for monitoring of the chromosome dynamics within single cells [12], extraction of chromosomes from cells [13, 14] and their subsequent isolation [15, 16], and probing the mechanical properties of chromosomes by translocation through nano-confined slits [17, 18] or using optical tweezers [19]. However, the controlled formation of defined chromosome ensembles by microfluidics has not yet been explored. A central challenge for the assembly of defined groupings of selected chromosome is that existing techniques for chromosome generation and isolation yield only limited numbers for each individual component. Because established methods for chromosome extraction from live cells suffer from low yield [14], and the de novo synthesis of human artificial chromosomes is further constrained by poor efficiency [20–23], tools used for downstream manipulation of the resulting molecule assemblies must be designed to minimize chromosome loss. To this end, microfluidic platforms similar to those developed to manipulate individual cells [24, 25] offer potential for the efficient formation of chromosome ensembles. In particular, droplet microfluidics [26] provides a straightforward approach for isolating single biological particles, with the number of components within each droplet defined by a Poisson distribution [27, 28]. Specific control over individual droplets for higher level manipulation of the discretized volumes, such as trapping, storing, releasing, metering, or merging [29, 30] can be achieved by appending additional microfluidic components following droplet formation, such as integrated microvalves [31–35], traps [36–40], or a combination of both [41–45] to yield the desired functionality. However, existing droplet microfluidic platforms are not designed with the goal of enabling efficient assembly of macromolecular sets with precise control over numbers and types within the resulting population.

Here we report an integrated droplet-based microfluidics platform designed to form user-specified chromosome ensembles while minimizing chromosome loss during the assembly process. The system combines dynamic microvalves for low volume on-demand chromosome sample plug injection, an array of individually addressable membrane displacement trap (MDT) elements [41] supporting a flexible set of droplet operations, and a multi-droplet merging zone to assemble selected chromosomes within a single aqueous volume. The platform provides automated control over multi-sample injection together with sample discretization, metering, and merging, allowing for the formation of an arbitrary set of chromosomes with single macromolecule resolution through a sequence of flexible operations that serve to maximize the utilization of the limited chromosome samples. Automated operation of the overall system is provided through a programmable microcontroller that coordinates all functional steps for operation of the integrated system. The platform is demonstrated through the controlled formation of a triplex set of paired human chromosomes, followed by isolation of the resulting ensemble within a single droplet suitable for downstream cell delivery.

2. Methods

2.1. Microfluidic chip fabrication

The devices were fabricated using a two-layer polydimethylsiloxane (PDMS) (Sylgard 184, Dow Corning, USA) process by standard soft lithography techniques. Briefly, two master molds for the fluidic and pneumatic layers were obtained by spin-coating negative photoresist (SU-8 2075, Microchem Corp, Newton, MA) at 3500 rpm for 60 s to construct 40 μm thick structures on a silicon wafer. The fluidic layer with 60 μm wide channels was formed by molding a mixture of PDMS prepolymer and curing agent at 10:1 ratio on the SU-8 pattern, partially curing at 80 °C for 30 min, peeling off from the master mold, and then punching to introduce access ports into the fluidic layer. The control layer was formed by spin-coating the PDMS prepolymer and curing agent at 20:1 ratio at 1500 rpm for 1 min, resulting in a 36 μm thick elastomer membrane above the channels, and partial cured at 80 °C for 10 min. The two layers were aligned and mated with the pneumatic layer still attached to its mold and fully cured at 80 °C overnight before being peeled off from the master mold, punched to introduce ports into the control layer, and bonded to a glass slide using oxygen plasma treatment for 60 s. To enhance hydrophobicity following the plasma treatment, the chips were stored at 115 °C overnight before use.

2.2. Experimental setup

The on-chip elastomer membrane microvalves were operated by individual off-chip pneumatic solenoid valves (Clippard Instrument Lab Inc., Cincinnati, OH) switched using an Arduino microcontroller. Pressurized air was supplied to the solenoid valves through a regulator (5–50 psi range, McMaster-Carr, Elmhurst, IL) to actuate the microvalves. The interface with the microchip was achieved through the flexible Tygon microbore tubing (0.51 mm ID, 1.52 mm OD, Cole-Parmer, Vernon Hills, IL) connected to 22 gauge needle segments (Hamilton, Reno, NV) inserted in the on-chip pneumatic/hydraulic access holes. The immiscible oil phase consisted of light mineral oil with Span-80 surfactant at 0.01% concentration (Sigma-Aldrich, St. Louis, MO), and the dispersed aqueous phase consisted of samples with sorted chromosomes in buffer. Mineral oil was employed as the continuous phase so that the final droplet containing the desired set of chromosomes may be readily converted into a giant lipid vesicle by the inverted emulsion technique [46]. For device characterization and demonstrations, the dispersed aqueous phase consisted of deionized water with dissolved red, blue, and green food dyes. To introduce the liquids into the chip, they were stored in custom plastic tubes and driven by application of a constant pressure using a regulator (0–10 psi range, Marsh Bellofram Co., Newell, WV). A custom off-chip housing with multiple inlets and one outlet manufactured by 3D printing allowed sequential introduction of multiple dispersed aqueous phase samples into the chip. A schematic diagram of the experimental setup for introducing the liquids into the chip is shown in figure 1(a) and for controlling the pneumatic microvalves using an Arduino microcontroller is shown in figure 1(b). A schematic diagram of the microfluidic chip is also illustrated in figure 1(c).

Figure 1.

Overview of the programmable sample plug formation, droplet storage, and droplet merging system. (a) A 3D printed housing connecting multiple pressurized vials to the chip is used for introducing oil and aqueous samples into the chip inlet. (b) On-chip microvalves are individually actuated using a microcontroller, allowing the entire procedure for sample injection, droplet generation, droplet release, and merging to be automated. (c) The device schematic showing the U-shaped channel to introduce samples into the chip, addressable PCAR storage unit with an array of MDT elements, and dynamic merging zone for trapping and combining multiple droplets. The fluidic layer is shown in blue and control layer in green.

2.3. Chromosome preparation and fluorescence imaging

Human fibroblast cells (BJ cells) were obtained from the American Type Culture Collection (Manassas, VA) and cultured in Eagle’s Minimum Essential Medium (Thermo Fisher Scientific, Waltham, MA) with 10% Fetal Bovine Serum (Thermo Fisher Scientific) at 37 °C in 5% CO₂. Once cells reached approximately 70% confluency in a T175 flask, Colcemid (10 μl ml⁻¹) (Sigma-Aldrich, St. Louis, MO) was added for 5 h. Mitotic shake-off was performed on the cells followed by incubation in hypotonic solution (0.075 M KCl) for 25 min at room temperature. Cells were centrifuged at 400 g for 5 min. The supernatant was removed, and cells were suspended at 5 × 10⁵ cells ml⁻¹ concentration in ice-cold polyamine buffer for 15 min on ice. The cells were vortexed for 1 min at high speed. To visualize whether chromosomes were successfully released, a mix of suspension and DAPI was examined under a fluorescent microscope. The released chromosomes were centrifuged and resuspended in RPMI media for a final concentration of approximately 4.5 × 10⁶ chromosomes per ml. The suspension was divided into three tubes and labeled separately with different fluorescent dyes including SYTO 12 green, SYTO 82 red, and SYTO 40 blue. Fluorescently labeled chromosomes were monitored using an inverted fluorescence microscope (Nikon TE2000) equipped with a 60x oil-immersion lens and a high sensitivity camera.

3. Results and discussion

3.1. Device design

As shown in figure 1(c), the integrated microfluidic device is composed of three functional units including a plug generator unit (valves 1–2), droplet storage unit (valves 3–6), and merging unit (valves 7–10). In the plug generator unit, a U-shaped inlet and a T-junction integrated with two pneumatic microvalves are used to disperse aqueous-phase sample to produce on-demand sample plugs. The sample plug injected into the main channel is then carried by the immiscible oil phase into a passive capture, active release (PCAR) storage unit composed of a linear array of MDT elements. Each MDT trap is integrated with a pneumatically controlled elastomer membrane that enables addressable processing of individual droplets by directly modulating the trap volume. A waste outlet is included in the design to allow residual sample to be shunted out of the system upstream of the storage unit. The storage unit itself is composed of a set of traps that allow for passive capture of incoming aqueous-phase droplets and active release of selective droplets using the integrated elastomer membrane that serve to directly modulate the trap volume. Stored droplets can be incubated and optically monitored in an immobilized state. After monitoring, the droplets can be released sequentially in a controlled manner. The released droplets move downstream, with selected droplets delivered to the merging unit, while undesired droplets are shunted to the waste outlet. Multiple aqueous-phase samples can be sequentially introduced, and their generated droplets are retrieved in a same manner. In the merging unit, the incoming droplets coalesce and fuse together to produce an ensemble of the desired contents and compositions in a final combined droplet.

3.2. System operation

Arduino code was developed to automate and synchronize the entire procedure for sampling and droplet manipulation. The developed code allows the user to initiate individual operations including sample injection and droplet generation, sequential droplet release, transferring droplets into the merging zone or shunting to the waste line, merging multiple droplets, and release of the final merged droplet. Referring to figure 1, two microvalves are used to inject a selected sample volume, ranging from nanoliter plugs to picoliter droplets, into the main channel. After formation of an array of immobilized droplets, the integrated trap membranes are selectively actuated to eject the droplets, allowing the desired droplets to be transferred into the merging zone while unwanted droplets are shunted to waste. Two coupled microvalves are used to control the flow into the trapping zone or waste outlet. This procedure is repeated for multiple samples, providing digital control over the droplet content delivered into the merging zone. In our experiments, chromosomes are captured in the PCAR storage and optically interrogated to identify an MDT trap containing a droplet with a pair of chromosomes. If no traps in the storage met the criteria, all traps are flushed to waste and new sample volumes are captured. Once identified, the desired droplet is parked in the merging zone while the capture process is repeated for the next chromosome sample. The final merged droplet containing the desired chromosome ensemble is finally released for off-chip collection. The entire procedure of device operation for each sample including sample injection and droplet generation, fluorescence monitoring of the droplet array, transferring of desired droplets to merging zone and shunting of undesired droplets to the waste can be accomplished in less than 1 min. Details of each process step for forming a desired chromosome ensemble are discussed below, and shown graphically in figure 2. Execution of each process step is depicted in the Supplementary Material (Video S1).

Figure 2.

Operational flowchart describing the sequence of sample injection, discretization, stationary droplet monitoring, metering, release, and merging for controllable chromosome ensemble formation.

3.2.1. Sample delivery

Before initial sample delivery, the entire microfluidic system is first primed with the immiscible phase. An off-chip 3D printed manifold is used for introducing multiple samples into the microfluidic chip (figure 1(a)). The manifold is connected to reservoirs for each chromosome sample as well as a reservoir containing rinse buffer, allowing the microfluidic system to be flushed between each sample injection event. Manual valves between the manifold and reservoirs are used to select the active fluid source. Each chromosome sample was stored in a custom plastic tube and driven through the microfluidic by application of a constant inlet pressure. Sample delivery to the on-chip injector was achieved using a wide (600 μm) U-shaped delivery channel. A microvalve positioned over the injection channel controllably gates sample into the main channel upon actuation through a T-junction for on-demand plug generation. The generated plug volume depends on the aqueous/oil inlets pressure ratio and valve opening time [37].

3.2.2. Sample discretization

The generated nanoliter sample plug, with approximately the total trap volume, is passively self-discretized into the traps. During this process, a portion of the sample plug fills each trap, but does not progress through the narrow trap outlet due to high Laplace pressure [47]. When the tail of the plug passes each trap entrance, it breaks off from the trapped fluid, leaving a droplet inside the trap with a size determined by the trap volume. In this way, an array of droplets with consistent size (458.8 ± 8.3 pl measured across a set of 40 droplets) is reliably formed. The immobilized droplet array enables optical monitoring and on-chip incubation of the parked droplets for an extended period. Alternately, the droplet array may also be formed by capturing individual picoliter droplets generated by the sample injector. In this mode of operation, once a droplet enters a given trap, the hydraulic resistance through the trap increases and the following droplets are directed to the downstream traps. The former approach was employed here as it offers reliable, rapid, and repeatable self-discretization with a low droplet size variance. The resulting droplets remained stable with no observable changes in size for at least 1 h.

3.2.3. Droplet retrieval and merging

The PCAR storage unit supports selective release of full or partial droplets from the array. This functionality is enabled using the integrated membrane positioned over each trap element. Application of pressure into the pneumatic control channel deflects the elastomer membrane into the trap, displacing the droplet from the trap. At sufficiently high pressure, full release of the droplet occurs, while lower pressures in smaller membrane deflections and partial metering of droplets from the trap. Unlike prior microfluidic systems designed for deterministic particle sorting [48–50], this approach allows for controlled reduction of discretized sample volume, providing the ability to reduce the number of chromosomes within a given droplet by ejecting a portion of the trapped volume. The merging unit enables sequential merging events required for multistep droplet processing. It is composed of a chamber for trapping multiple droplets, with two parallel bypass flow paths and two integrated microvalves. One microvalve serves to open the merging zone or route fluid through the bypass channels, and the second microvalve directly beneath the merging chamber facilitates active merging between the surfactant-stabilized droplets by rapidly oscillating the chamber volume. This microvalve also acts as an MDT membrane to eject the final merged droplet. The chamber size of 260 × 320 μm is designed to accommodate up to six 0.5 nL droplets for merging.

3.3. System characterization

3.3.1. Sample discretization

The ratio of hydrodynamic resistance between each trap and its associated bypass path is the primary parameter that dictates performance of the self-discretization process. The trap is designed such that the water/oil interface must slightly deform in order to enter the trap with a narrow entrance width, serving to avoid undesired coalescence between a stored droplet and sample flowing within the main channel. Competition between hydrodynamic resistance in the bypass channel and a combination of hydrodynamic resistance in trap together with Laplace pressure as the water/oil interface deforms during trap filling determines whether sample will progress through the trap or the bypass channel. Furthermore, because hydrodynamic resistance depends on fluid viscosity, self-discretization performance also depends on the sample plug volume. The passive trapping technique has been previously studied [51–53], including analytical modeling of the process based on evaluation of the ratio of hydraulic resistance between the flow path through a single trap (R_trap) and through the bypass channel around the trap (R_bypass) [51]. Using an adaptation of this model, with the hydrodynamic resistance in each branch determined using an approximation based on Poiseuille flow in a rectangular channel [54, 55], figure 3(a) presents a phase diagram depicting successful vs. incomplete self-discretization events when capturing a sample plug within the four-trap array over a range of hydraulic resistance ratio (R_trap/R_bypass) and ratios of sample plug volume to the total volume of the full trap array. For the device design used in this work, with a trap diameter of 100 μm, entrance width of 55 μm, entrance length of 35 μm, bypass channel width of 60 μm, and bypass length of 1050 μm, reproducible sample self-discretization is achieved within all traps when employing large sample plug volumes or when using designs with low hydrodynamic resistance ratios.

Figure 3.

Phase diagrams showing (a) the impact of injected sample volume and ratio of hydrodynamic resistance between the trap and bypass channel on passive discretization performance of a sample plug with initial volume (V_sample) into an array of traps (total volume 4 × V_trap), and (b) the impact of initial droplet volume (V_droplet) and applied MDT membrane pressure on droplet release at low continuous phase flow rate (<0.5 nl s⁻¹) within the main channel.

3.3.2. Droplet ejection and metering

Upon pressurizing the pneumatic control channel for a given trap, three distinct operational scenarios may occur when operating at a low (<0.5 nl s⁻¹) continuous phase flow rate within the main channel: no droplet release, intact droplet release, and droplet splitting. The relevant regime was found to depend on the applied membrane pressure as well as the initial volume of the droplet within the trap. The behavior under each regime is presented in figure 3(b). In the first scenario, the applied membrane pressure is not sufficient to displace the droplet out of the trap, and no release occurs. In the second regime, deflection of the membrane displaces the entire droplet into the main channel, allowing it to be removed from the trap by the continuous phase flow. In the third scenario, extreme deflection of the membrane imparts sufficient force to overcome cohesive surface tension during ejection, leading to droplet splitting.

As the continuous phase flow rate increases, the high strain rate imposed on the emerging sample volume can lead to sufficient shear stress to generate a secondary droplet. This process can be used to controllably meter sample volume from the trap by partially ejecting the trapped volume at continuous phase flow rates above 0.5 nl s⁻¹. As revealed in figure 4(a), the volume of the metered droplet may be reliably controlled by adjusting the applied pressure on the MDT membrane (figure 4(b)) for a given continuous phase flow rate in the main channel. For example, at a nominal flow rate of 0.53 nl s⁻¹, a 950 pl droplet was sequentially metered in two steps with average volumes of 445 pl metered at each step, allowing the final droplet remaining within the trap to be reduced to approximately 10 pl, while finer control over the ejected volume was achieved at higher flow rates. The metering process offers a simple method to remove a portion of the trapped sample volume, allowing the contents to be tuned following the initial trapping step.

Figure 4.

(a) Ejected sample volume for a multi-step metering sequence performed using an MDT membrane under different continuous phase flow rates, and (b) the corresponding pressures applied to the MDT membrane for each metering step.

3.4. Chromosome ensemble formation

Solutions containing three differentially labelled chromosomes were sequentially injected into the chip and discretized into the traps, with the number of macromolecules present in each trap following a Poisson distribution [27, 28]. In these experiments, a droplet volume of approximately 450 pl was employed for each trapping event. At the selected chromosome concentration (4.5 × 10⁶ chromosomes ml⁻¹), the probabilities for a single droplet to contain a given number of chromosomes are approximately 14% for 0 chromosomes, 27% for 1 chromosome, 27% for 2 chromosomes, 18% for 3 chromosomes, and 14% for 4 or more chromosomes. In our experiments, within the 48 tested droplets, 8 droplets had 0 chromosomes, 16 droplets had 1 chromosome, 12 droplets had 2 chromosomes, 8 droplets had 3 chromosomes, and 4 droplets had 4 or more chromosomes. Thus, the measured distribution probabilities for each case were approximately 16.6%, 33.3%, 25%, 16.6%, and 8.3%, respectively (figure 5(a)). After each discretization event, the droplet array was monitored by fluorescence microscopy to identify traps capable of yielding a pair of chromosomes. When a single trap contained the desired chromosome pair, e.g. figure 5(b)(i) the droplet was directly ejected. When no single droplet contained a chromosome pair, but more than two chromosomes were present in a trap, e.g. figure 5(b)(ii) the droplet was metered in steps to yield a droplet with paired chromosomes. Similarly, if a pair of droplets with single chromosomes was found, e.g. figure 5(b)(iii) both droplets were ejected. Finally, if no suitable droplet combination was generated, the traps were actuated to eject the droplets to waste, and a new set of droplets was captured. In each case the desired droplets were moved into the merging zone, while all other droplets were shunted to waste, and the process was then repeated for the next chromosome type. In total, droplets composed of three differentially labeled chromosome pairs were combined into a single droplet in the merging zone to form a spatially confined genome, as shown in figure 5(c). Execution of the full procedure including passive discretization into the trap array, fluorescence monitoring, chromosome injection, droplet transport into the merging zone, and shunting of unwanted droplets to waste required approximately 45 s for each chromosome solution. Chromosome structure and stability was monitored visually throughout each experiment, with no morphological changes noted.

Figure 5.

(a) Comparison of theoretical and experimental chromosome distributions in the microfluidic trap chip. (b) Images of the trap array following three discretization events, in which (i) at least one droplet containing a single chromosome pair is present, (ii) a metering step is needed to yield a droplet with a chromosome pair, and (iii) a pair of droplets encapsulating single chromosomes is selected for merging. (c) Fluorescent images of three different pairs of chromosomes, together with a composite image revealing the final chromosome ensemble.

4. Discussion

The number of traps used in this study was selected as a tradeoff between device complexity, operational efficiency and speed, and optimal chromosome utilization. The design containing four traps was found to be sufficient to ensure that the vast majority of discretization events yielded a combination of traps with at least two chromosomes, either within an individual trap or across multiple elements, allowing for selection of a single pair through at least one of the operations outlined in figure 5. Increasing the number of traps would add system complexity without enhancing either capture efficiency or processing speed, and thus was not explored. Conversely, a smaller number of traps would serve to improve chromosome utilization for a single discretization event by reducing the number of chromosomes shunted to waste. However, this approach would require multiple discretization events to generate the desired outcome, reducing system throughput.

A practical limitation of the current device design is the use of a single sample delivery channel. The wide U-shaped channel design allows rapid delivery of chromosomes to the downstream MDT array using a constant pressure source from the inlet manifold, while also preventing air bubbles or residual sample from entering the main channel during loading, but the large dead volume of the channel and associated interconnects results in significant sample loss when changing between each chromosome sample. To fully take advantage of the efficient chromosome utilization offered by the active trapping elements, individual inlets should be employed to allow ensembles to be constructed without flushing residual sample from the delivery channel between each chromosome type.

We also note that the 60× oil immersion lens employed for chromosome imaging did not provide sufficient resolution to distinguish chromosomes solely by karyotype without staining. While chromosomes with large size variations can potentially be directly differentiated without staining using the existing configuration, the ability to identify all chromosomes by banding or morphology alone would require higher magnification optics.

5. Conclusion

This work reports a multifunctional microfluidic system for programmable discretization and manipulation of picoliter sample volumes and demonstrates its application to the unique challenge of constructing a defined ensemble of paired human chromosomes. The method leverages integrated membrane valves for versatile execution of complex functionalities including controllable sample injection, on-demand sample droplet and plug generation, self-discretization, and metering, as well as reliable release and sequential merging of selected droplets to collect a controlled number of encapsulated macromolecular contents. The operation of the microfluidic platform is optimized through the evaluation of relevant physical and operational parameters, and automated through a simple microcontroller interface to enable the selective isolation and direct manipulation of single chromosomes and their assembly into a defined genome ensemble. The combination of stochastic chromosome encapsulation and flexible deterministic manipulation provided by the microfluidic platform can serve as a unique tool for preparing complex sets of chromosomes for cell delivery, and ultimately for the assembly of isolated chromosomes into a functional cell nucleus. This latter goal will require the formation of a nuclear envelope by incubation of mitotic nuclear envelope fractions with interphase cytosol as has been done in other in vitro re-assembly systems, serving to move toward the controlled biogenesis of synthetic mammalian cell nuclei with customized genomes.

Data availability statement

The data that support the findings of this study are available upon reasonable request from the authors.