Dual-fiber microfluidic chip for multimodal manipulation of single cells

Abstract

On-chip single-cell manipulation is imperative in cell biology and it is desirable for a microfluidic chip to have multimodal manipulation capability. Here, we embedded two counter-propagating optical fibers into the microfluidic chip and configured their relative position in space to produce different misalignments. By doing so, we demonstrated multimodal manipulation of single cells, including capture, stretching, translation, orbital revolution, and spin rotation. The rotational manipulation can be in-plane or out-of-plane, providing flexibility and capability to observe the cells from different angles. Based on out-of-plane rotation, we performed a 3D reconstruction of cell morphology and extracted its five geometric parameters as biophysical features. We envision that this type of microfluidic chip configured with dual optical fibers can be helpful in manipulating cells as the upstream process of single-cell analysis.

I. INTRODUCTION

Cell research plays an important role in disease diagnosis and investigating cancer metastases.^1,2 Single-cell manipulation^3,4 is imperative in biotechnology to facilitate many potential applications, such as cell injection,⁵ imaging,^6,7 positioning,⁸ and electrical properties characterization.⁹ Basic manipulation includes translation and rotation of cells, among many others. Rotational manipulation^10–13 has been more challenging than translation. Over the years, different approaches have been practiced for rotational manipulation at the single-cell level,^14–16 mainly including mechanical, hydrodynamic, optical tweezers, magnetic, acoustic, and electrical means. The mechanical methods^17,18 use micromanipulation devices to precisely control a probe (e.g., glass pipette) to realize the 3D rotation of single cells. However, they require direct contact with cells and thus easily damage them. The hydrodynamic methods realize the micro-recirculation of solution through the special microstructure, thus making the cells rotate, but this method relies on complex microstructures.^19,20 The magnetic methods^21,22 use external magnetic fields to perform 3D cell rotation and require a cell sample pretreatment. Acoustic methods^23–25 use vibration principle to realize the 3D rotation of single cells, and the vibration modes depend on the microchannel structure.²⁶ Electrical methods^27–29 are based on the principle that cells can be polarized and moved in the alternating electric field to realize 3D rotation. Electro-thermal and electro-osmotic effects may potentially damage cells. Optical methods commonly include optical tweezers^30,31 and optical fibers^32,33 to achieve single-cell rotation. Optical tweezers use a low-power laser beam to exert trapping force and torque onto cells and cause them to rotate. Utilizing multiple optical tweezers can achieve out-of-plane rotation of single cells.³⁴ Compared with traditional optical tweezers, the optical fiber method can realize the rotation of single cells simply by two counter-propagating and misaligned optical fibers. Recently, we used two fibers and four 3D electrodes to achieve single-cell multi-parameter measurements,³⁵ but there are still some limitations in the application of cell study, for example, only in-plane rotation can be achieved.

In this paper, we present microfluidic chips configured with two counter-propagating fibers that can achieve multimodal manipulation of single cells. When the two fibers are aligned, single cells can be trapped and stretched by laser-induced forces. If there is a certain offset between the two optical fibers, multimodal manipulation of single cells can be realized, including spin rotation and/or orbital revolution, in-plane or out-of-plane. We successfully reconstructed the 3D morphology of single cells from a stack of images from out-of-plane spin rotation and calculated their physical parameters from the 3D image. This 3D imaging technique provides a possible alternative to confocal microscopy with no need for labeling, as a new tool for cell characterization. The dual-fiber microfluidic chips can efficiently realize single-cell multimodal manipulation and provide a convenient platform for single-cell research.

II. MATERIALS AND METHODS

A. Working principle

The working principle of multimodal manipulation of a cell is illustrated in Fig. 1, based on light-induced forces exerted on the cell. Generally, the change of light momentum caused by light irradiating the cell will induce axial and gradient forces in its axial and normal direction.^36,37 The axial force is generated by the photon hitting the cell along the propagation direction of the light, which is generally called scattering force; the gradient force is caused by the uneven intensity of the light field, and its direction is perpendicular to the direction of light propagation.

FIG. 1.

Working principle of double optical fibers embedded in microfluidic chips for multimodal manipulation of single cells. (a) Optical trap. (b) Optical stretch. (c) Optical translation. (d) Orbital revolution. (e) In-plane spin rotation. (f) Out-of-plane spin rotation. Note that P₁ or P₂ is the laser power per fiber and d is the cell diameter.

Visible light is not suitable for manipulating cells because of two reasons. First, the force generated by visible light is one or more orders of magnitude smaller than the cell weight and fluidic force. Second, heating of the solution via light absorption may cause thermal damage to cells. Thus, near infrared single-mode laser is often used as the light source. The gradient force and axial force exerted on cells are related to cell volume and the laser wavelength. When the cell radius is much larger than the laser wavelength, which is normally the case, the forces can be computed through a ray-optics approximation.^38,39

The optical field of a non-focused Gaussian beam can be described by

$$w(z)=\sqrt{1+{\left({\displaystyle \frac{\lambda z}{n\pi w_0^2}}\right)}^2},$$ (1)

where w is the beam waist, z is the distance from the fiber end, λ is the laser wavelength in vacuum, n is the index of refraction of the solution, and w₀ is the beam waist as the beam exits the fiber.

The optical forces acting on the cell are obtained by applying the relations for the scattering and gradient components exerted by each ray^38,39 via

$$F_s={\displaystyle \frac{nP}{c}\times \left\{[1+R\cos 2\theta ]-{\displaystyle \frac{T^2[\cos (2\theta -2\gamma )]+R\cos 2\theta }{1+R^2+2R\cos 2\gamma }}\right\},}$$ (2)

$$F_G={\displaystyle \frac{nP}{c}\times \left\{[1+R\sin 2\theta ]-{\displaystyle \frac{T^2[\sin (2\theta -2\gamma )]+R\sin 2\theta }{1+R^2+2R\cos 2\gamma }}\right\},}$$ (3)

where P is the power carried by each ray, c is the light velocity, R and T are the Fresnel reflection and transmission coefficients, respectively, at the cell surface at a given incidence angle θ, and γ is the refraction angle.

Denote the misaligned offset of the two identical counter-propagating optical fibers by L, and the cell motion patterns are categorized by the value of L. When L = 0, the total force on the cell was zero, and the cell was stably trapped. This results in an optical trap [Fig. 1(a)] that drags the cell into the center point. When the trapping force is increased by increasing the laser power on the cell surface such that it is greater than the cellular mechanical strength, the cell can be stretched and deformed in situ^40,41 [Fig. 1(b)].

When L ≠ 0, which may be easily the case in reality, cells can exhibit several patterns. If L > d (the cell diameter), cells can be only subject to the scattering force of one optical fiber and be translated toward the other optical fiber [Fig. 1(c)]. If L < d, the scattering forces from each laser beam are not collinear and can generate a torque on the trapped cell, causing it to rotate.^42,43 Specifically, when d/2 < L < d, the cell does orbital revolution, as shown in Fig. 1(d); when 0 < L < d/2, the cell does spin rotation, as shown in Figs. 1(e) and 1(f). Depending on which plane the misalignment occurs, the rotation can be in in-plane rotation [Fig. 1(e)] or out-of-plane rotation [Fig. 1(f)].

B. Chip design and fabrication

Figure 2(a) shows the sketch of the microfluidic chip. In order to facilitate the integration of the optical fibers into the microchannel, a 125 μm-thick microchannel was fabricated by photolithography. Constrained by the microchannel geometry, two counter-propagating optical fibers of 125 μm in diameter were inserted after bonding of the glass substrate and polydimethylsiloxane (PDMS) microchannel. The optical fibers were carried by a 40-nm-resolution 3D micromanipulator (MP285, Sutter, USA) to adjust the spatial position. The y-axis misalignment was confirmed by direct microscopic observation, and the relationship between the optical coupling power and misalignment was recorded and used to adjust the z-axis misalignment when needed. To prevent possible fluidic leakage through the gap between the round optical fiber and the square channel, we sealed the outside of the channel with glue. The position of the optical fibers is fixed and non-adjustable after the glue is cured, multiple chips with different optical fiber misaligned offsets were made to demonstrate the multimodal manipulations of cells. Note here that the optical fibers are intentionally placed in the downstream of the main microchannel, which runs the sample medium and two sheath flows. The sheath flows are used to focus the cells to the centerline of the main channel to ensure that cells enter the effective zone of the optical fibers.

FIG. 2.

The microfluidic chip with two counter-propagating optical fibers and the manipulation experimental setup. (a) The sketch of the microfluidic chip with two optical fibers orthogonal to the main channel. (b) The fabrication process of the microfluidic chip. (c) The real picture of the microfluidic chip. (d) Illustration of the experimental setup.

In principle, we can obtain all the manipulation modes (Fig. 1) of the cell if we can adjust L in the fly. This can be achieved by adding some on-chip design, which can change the position of one or both of the optical fibers as the proper actuation in association with external stimulation. For example, adding a pneumatic channel alongside the fiber-embedded channel may change the y-axis misalignment. This is not the focus here, but the concept of multimodal manipulations can be demonstrated by current experimental settings.

Figure 2(b) illustrates the fabrication procedure of the device. (i) The mold of the microchannel was made of SU-8 and fabricated by soft-lithography technique. The microdevice contains the main flow microchannel and the fiber channel. (ii) PDMS was poured on the SU8 mold to replicate the microchannel. (iii) The solidified PDMS was peeled off from the mold. (iv) The PDMS and the glass substrate were treated with oxygen plasma and bonded together. The assembled microfluidic device was baked at 120 °C for 1 h to further improve bonding strength ultimately. (v) Two optical fibers were inserted into the fiber microchannel. (vi) Glue was fixed at the junction of the optical fiber and PDMS microchannel in order to avoid leakage. The photo of the fabricated device is shown in Fig. 2(c).

C. Experimental setup

Figure 2(d) illustrates the experimental setup. The laser with a wavelength of 980 nm provided by a laser transmitter (VLSS-980-B, Connect Fiber Optics, China) was used in the experiments. Such a wavelength is known to be poorly absorbed by water,⁴⁴ and thus the heat damage to the cells will be minimal. The laser beam is split evenly into two single-mode optical fibers (HI 1060, Corning) via a 1 × 2 fiber coupler (Gould Fiber Optics). Cell solution was loaded to the microchannel by withdrawing the pump (Legato 200, KD Scientific). The microfluidic chip was placed on an inverted microscope (Nikon ECLIPSE Ti-U), and a camera (Nikon DS-Ri1) working at 30 Hz was mounted on the microscope for image capture and video recording.

D. Cell preparation

Normal human breast cells MCF10A were used in this experiment and obtained from School of Medicine of Tsinghua University. The cells were cultured using an incubator (Forma 381, Thermo Scientific, USA) at 37 °C in 5% CO₂. The culture medium was high-glucose Dulbecco's Modified Eagle's Medium (DMEM, Life technologies, USA), supplemented with 10% fetal bovine serum (FBS, Life technologies, USA) and 1% penicillin-streptomycin (Life technologies, USA). Cells were rinsed with phosphate buffered saline (PBS, pH 7.4) twice and lifted off by treating with trypsin for 5 min. The cell suspension was washed three times by centrifuging at 300 g for 5 min, removing the supernatant with a pipette, and re-suspending the cell pellet in 1 ml PBS buffer supplemented with 1% BSA (Bovine Serum Albumin) to avoid adhesion and shaken gently to obtain the uniform cell suspension. To facilitate single-cell loading, the cell suspension was then diluted with additional PBS until a density of 1.5 × 10⁵ cells per ml was achieved.

III. RESULTS AND DISCUSSION

A. Optical trap, stretch, and translation

When the dual fibers are well aligned, cells can be trapped in the center point of two microchannels and stretched in situ. To estimate the distribution of the optical forces exerted on a single cell, a finite element simulation (COMSOL Multiphysics 5.5) was performed to show the electric field distribution generated around the dual fibers. For a cell irradiated by an arbitrary monochromatic wave, the time-averaged radiation-induced forces acting on it can be derived from the nonrelativistic Lorentz force

$$F=\pi r^3{\epsilon }_{m}Re[K_{CM}]\mathrm{\nabla }{E}^2+{\displaystyle \frac{1}{2}{\epsilon }_{m}Im[K_{CM}]\left[{\sum }_l{E}_l^{\ast }\mathrm{\nabla }{E}_l\right],}$$ (4)

where ɛ_m is the permittivity of the medium, the subscript runs over l = x, y, z, and the superscript * signifies the complex conjugate. According to Eq. (4), the optical field force and the electric field distribution present a functional relationship. In the case that the linear coefficient cannot be accurately evaluated, the optical field force can be qualitatively estimated by simulating the electric field power.⁴⁵ The simulation results [Fig. 3(a)] show a wonderful symmetry of the electric fields around the fiber channel center, which leads to the spatial equilibrium of the cell both horizontally and vertically. Under the condition that the fluidic drag force is less than the trapping force, a cell flowing through the main channel would be trapped to the center point of the two microchannels, and also the initial position of the cell is not in the middle of the main streamline.

FIG. 3.

Cell trap, stretch, and translation. MCF10A cell samples were used in experiments. (a) The simulation results of the electric field in the designated region of the microfluidic channel, when two fibers are aligned well (i.e., L = 0). (b) One cell being trapped (100 mW per fiber) and (c) stretched (400 mW per fiber). (d) Two cells were being trapped (100 mW per fiber) and (e) stretched (400 mW per fiber). (f) The simulation results of the electric field in the designated region of the microfluidic channel, when two fibers are not aligned (L > d). (g) One cell being landed on the end of the other fiber after translation by one fiber.

The microfluidic device with well-aligned dual optical fibers was used to confirm the trap and stretch manipulation. According to Eq. (1), the diameter of laser spots at the center of the microchannel is about 12 μm. This size is comparable to the cells we used in experiments, satisfying the purpose of cell manipulations. Referring to the literature, the laser power per fiber was set to be 100 mW, and then the cell medium was flushed through the main microchannel by increasing the flow rate from zero to 100 μm/s. It was found that below 60 μm/s, the cells were nearly 100% trapped by overcoming the fluidic drag force. However, there was the increasing ratio of failure of trapping once the flow rate was increased, probably due to the higher fluidic drag force. The cell could be trapped even when the cell flowed in with some offset against the middle of the microfluidic streamline but its trajectory fell into the effective zone, which is defined as the overlapping area by the two laser beams. Once it entered the zone by flow, the cell was dragged with acceleration to the centerline of the aligned optical fibers under the gradient force, and, then, the cell was pushed to the center of the main channel by the scattering axial force. For the case where the two optical fiber ends were aligned well with the channel walls, the cell was centered adaptively and trapped steadily in the main channel. Normally, it took about 2 s for the cell to be trapped from the streamline to the trap site. Figure 3(b) shows the snapshot of the MCF10A cell stably trapped at the center point. Once the cell was stably trapped on site, the laser power can be increased. Since the scattering force on the cell is proportional to the light intensity, the cell can be stretched by increasing the laser power. Figure 3(c) shows that when the power per fiber was increased to 400 mW, the cell was stretched, and the deformation was about 27%. This dynamic deformation process can be made use for characterizing cellular mechanical properties, as one powerful application of the manipulation.³⁵

Interestingly, the well-aligned dual optical fibers were found to be able to trap multiple cells as well. When they are both in the effective zone, multiple cells will be trapped and form a cell chain under the constraints of scattering and gradient forces. Regardless of their initial distance, the two cells will eventually converge in the middle of the flow passage and join together to form a cell chain. If the laser power is strong enough, the optical trap can capture three or more cells. The number of cells or length of the cell chain is positively related to the laser power. Like a single cell, the cell chain can be stretched once the laser power is increased further. Under the action of the scattering force, the long axis direction of the cell chain was consistent with the direction of light propagation. Figures 3(d) and 3(e) show that two MCF10A cells were trapped and stretched as a cell chain. When the optical power was increased by four times to 400 mW, the deformation of the cell chain was 4.5%. Compared to the single cell, the cell chain had six times smaller deformation. This observation indicates that the cell chain has different mechanical and optical characteristics than single cells.

When the misaligned offset is greater than the cell diameter, the cell is only subject to the scattering force of one of the optical fibers and the net force in the horizontal direction translates the cell to the end of the counter-propagating optical fiber. Figure 3(f) shows the simulation results of the electric field distribution for the case of misaligned optical fibers. The microfluidic device with misaligned dual optical fibers (L > d) was used to demonstrate the translation process. When the optical power is more than 100 mW, the cell can be easily translated and finally land on the end of the other fiber. Since the distribution of flow rate in the microchannel is parabolic and the flow rate near the wall is minimized to 0, the fiber-end-landed cell is unlikely washed away. Under this circumstance, it is difficult to deform the cell by increasing the optical power, because the scattering force is too small at a great distance. Figure 3(g) shows the time-lapsed image of the translation trajectory of one MCF10A cell. As the scattering force declined significantly with the distance, the cell moving speed decreased gradually.

B. Orbital revolution and spin rotation

Cells were found to rotate with different rotation patterns when the misaligned offset was varying against the cell radius. Generally, orbital revolution and spin rotation were the two patterns, and the spin rotation was a particular case of the former. Orbital revolution plus spin rotation was observed when d/2 < L < d, and spin rotation was observed when 0 < L < d/2. Figure 4(a) shows the simulation results of the electric field distribution of the dual optical fibers for d/2 < L < d. The electric field distribution reveals that the scattering force generated by the dual optical fibers are not collinear. This generates a torque on the cell, causing it to self-spin. In the meantime, the scattering forces in the x axis and the gradient forces in the y axis generated by the two optical fibers would be location-dependent. The overall effect is always to generate a centripetal force pointing to the intersection between the two optical fiber cores and the middle streamline along the microchannel. This centripetal force drives the cell to exhibit the orbital revolution. Therefore, the cell has a hybrid rotation pattern overlapped by orbital revolution and spin rotation.

FIG. 4.

Rotation patterns of cells when the two fibers are misaligned in space. MCF10A cell sample and 200 mW per fiber were used. The simulation results of the electric field (a) and the experimental results of one cell (b) performing in-plane orbital revolution, when d/2 < L < d and L occurs in the y axis. The dotted blue ellipse depicts the trajectory of the cell center. The simulation results of the electric field (c) and the experimental results of one cell (d) performing in-plane spin rotation when 0 < L < d/2 and L occurs in the y axis. The curved arrow indicates the spin rotation direction. Note that the other two cells out of the effective zone do not rotate at all.

Figure 4(b) shows the time-lapsed images of the dynamic trajectory of one MCF10A cell exhibiting orbital revolution in the horizontal plane, in case that the misaligned offset was 15 μm along the y axis. In the beginning, the cell was positioned at the light axis of the right fiber. When it was repelled to the left fiber, the cell was decelerated in the x axis direction until the velocity reached 0. Then, it was accelerated by the scatting force generated by the left fiber and moved toward the right fiber. In the y axis direction, the cell was first accelerated by the gradient force of the right fiber in the y axis direction and then decelerated by the gradient force of the left fiber. When the velocity of the cell in the y-axis direction became 0, it was accelerated by the gradient force and moved toward the +y-axis direction (Video S1 in the supplementary material).

The cell would exhibit spin rotation in a dominant manner when the misaligned offset was further shortened. When the misaligned offset distance is 0 < L < d/2, the cell does spin rotation about the center of the flow channel. The scattering force will generate torque on the cell, causing it to rotate. Because the misaligned offset of the dual fibers is less than the radius of cells, the gradient force can prevent the cells from shifting and fix it at the center of the flow channel. Figure 4(c) is the simulation results of the electric field distribution of the dual fibers with misaligned offset 0 < L < d/2. Because the two beams are not collinear, the scattering force acting on the cell still produces a torque on the cell. The cell can be located in the center of the microchannel and does spin rotation. Figure 4(d) shows one MCF10A cell exhibiting spin rotation at ∼225°/s under the misaligned fibers acting in the horizontal direction (200 mW per fiber). The rotation speed is nearly proportional to the optical power and the direction of spin can be changed by shifting the misaligned offset between the dual fibers (Video S2 in the supplementary material).

C. In-plane rotation and out-of-plane rotation

In the above analysis, the misaligned offset occurs in the horizontal (y-axis) direction, and the rotation motion of the cell belongs to in-plane rotation; while when the dual fibers are misaligned in the vertical (z-axis) direction, the cell presents out-of-plane rotation. Because the density of cells is close to that of the solution, the buoyancy and gravity can be balanced for simplicity. Thus, similar to the in-plane rotation, the different misaligned offset in the vertical direction can also induce orbital revolution and spin rotation of cells. Figure 5 shows a sequence of images for out-of-plane spin rotation of a MCF10A cell. The cell was rotated at speed of 100°/s (Video S3 in the supplementary material). The speed was calculated by counting the time and angles for selected feature points on the cell, which was video recorded for several resolutions. Not surprisingly, multiple cells trapped in the center of the microchannel can also do out-of-plane rotation (Video S4 in the supplementary material). The direction of rotation can be controlled by changing the relative spatial position of the two optical fibers.

FIG. 5.

Out-of-plane spin rotation of the MCF10A cell when 0 < L < d/2 and L occurs in the z axis.

D. 3D reconstruction of cell surface via spin rotation

Enabled by out-of-plane rotation, a stack of cell contours can be imaged in several rounds and reconstructed to form the 3D morphology of the cell. In the experiment, we used a 40× objective to record the cell contour images. The recorded video clip was converted into individual image frames. Each gray-level image of one round was thresholded into a binary image. The threshold was set to segment sufficiently the cell contour from the surrounding background. Cell contour was extracted from the binary images and the contour points had their 3D coordinates. A standard alpha-shape algorithm^46,47 can be used to reconstruct the 3D morphology from these 3D contour points (Video S5 in the supplementary material). Then, we extracted the geometric parameters for the cell, like volume, surface area, roughness, and ellipticity from the 3D morphology.

Figure 6(a) illustrates the image processing steps for 3D reconstruction of MCF10A cell surface with 241 image frames recorded at the angular displacement of ∼1.5°. To eliminate the noise in contour points, in practice, we extracted contour points from several rounds. All these contour points in 3D formed a point cloud, from which the 3D surface was reconstructed consisting of optimized triangles. Geometric parameters of the cell were retrieved from the reconstructed 3D morphology to demonstrate the potential measurement application of 3D imaging. Figure 6(b) illustrates the 3D reconstructed models for ten MCF10A cell samples. The volume and the surface area of MCF10A cells were averaged as773.88 fL and 244.78 μm², respectively. To further study the geometry of cells, we fitted the 3D morphology cloud points as a tri-axial ellipsoid model, for which ellipticity is denoted by the ratio of the length of the long axis to the length of the short axis, and roughness is denoted by the root mean square of the deviations between the cloud points and the fitting ellipsoid model. The ellipticity and roughness of MCF10A cells were averaged as 1.2 and 0.25 μm, respectively. Figure 6(c) plotted the distribution of the five parameters after normalization for the ten cell samples. The results reveal the difference between cells, but the set of area and volume parameters can reflect more differences.

FIG. 6.

3D reconstruction of the cell surface. (a) The image processing steps of 3D reconstruction. A stack of images are captured at an interval of ∼1.5° for the rotation angle. These images are then binarized and the cell contour points in 2D are extracted. All these points are then projected into 3D to generate the cell surface. (b) The reconstructed 3D surface of ten MCF10A cell samples. (c) The distribution of five geometric parameters for the ten cell samples. Each color represents one cell sample. (d) PCA results of the five parameters in (c) for the ten cell samples.

Multi-dimensional information can better describe the characteristics of cells, but the problem may be redundant characterization information and difficult visualization in the high-dimensional information. In order to realize the visualization of cell distribution and remove redundant information, principal component analysis (PCA) can be used to reduce the high-dimensional data into two dimensions. To demonstrate this idea, the type-wise values of the five property parameters were processed by PCA, and the result is shown in Fig. 6(d). It can be seen that the ten samples can be separated mainly in the x axis, which is the first principal component, and can be further separated in the y axis, which is the second principal component. Actually, the corresponding weights of the first two principal components are 95.4% and 4.5%, respectively, which retained almost all the information of the original five-dimensional data. Unlike the original data, the reduced two-dimensional data now have no physical meaning but offer a more straightforward distribution. This technique can be helpful in pruning the features that characterize single cells or cell types.

IV. CONCLUSIONS

We achieved multimodal manipulation of single cells on chip by tuning the misaligned offset of dual optical fibers orthogonal to the main channel. The microchip is capable of not only trapping and stretching single cells but also rotating cells in spin and/or orbital revolution, in a 2D or a 3D manner. When aligned well, the optical fibers can trap and stretch single cells. When misaligned, the optical fibers can cause spin rotation and orbital revolution of cells in the horizontal (in-plane rotation) and vertical (out-of-plane rotation) direction, simply by adjusting the misaligned offset in value and direction. Taking advantage of out-of-plane rotation, the 3D morphology of single cells can be reconstructed, from which multi-parameters of physical properties can be calculated. The flexible configuration of optical fibers and resultant manipulation patterns are expected to be a powerful tool in single-cell analysis.

SUPPLEMENTARY MATERIAL

See the supplementary material for (1) orbital rotation (Video S1), (2) spin rotation (Video S2), (3) out-of-plane rotation of single cell (Video S3), (4) out-of-plane rotation of two cells (Video S4), and (5) 3D reconstruction (Video S5).

AUTHORS’ CONTRIBUTIONS

L.H. and Y.F. contributed equally to this work.

DATA AVAILABILITY

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