Image-based particle tracking methods have numerous applications in biological studies at the cellular and molecular level. For example, tracking thermal motions of micron-sized spherical particles embedded in the cell cytoplasm reveal microheterogeneity of cellular rheology (1). Fluorescent microspheres conjugated with protein analogs as well as colloidal gold beads have been used to study mobility of lipids and proteins in the cell membranes (2) and to track motions of kinesin molecules on microtubules (3). Improved instrumentation and multiple particle tracking algorithms now permit high-throughput, nanometer-scale positional resolution using time-lapse differential interference contrast (DIC) (4,5) and fluorescence imaging (4).

However, there remain impediments to accurate tracking at the nanometer scale. Despite advances in motorized stage design, stages still exhibit lowfrequency positional drift due to thermal cooling of stepper motors and high-frequency vibrations arising from relative motion of sample and imaging instrumentation [e.g., charged-coupled device (CCD) cameras]. Although immobilized beads adhered to glass coverslips have been used as stationary markers (6), it is difficult to control their precise location and stability during cell culture. Furthermore, any fiduciary markers confined to the coverslips will not be in focus when imaging just a few microns away and are thus unsuitable for correcting tracking errors at the tops of cells. Similarly, immobilizing large beads is not feasible since they have small attachment area to the glass surface and can easily become dislodged. Other avenues to mitigate mechanical instability are expensive piezo stages, highperformance vibration isolation tables, and remote cooling of camera CCD chips. Even with these improvements, there will always remain some uncertainty as to whether particle trajectories represent true motion or coverglass motion artifacts.

To overcome uncertainty of particle position at both high and low frequencies (e.g., vibration and stage drift, respectively) and to enable tracking of coverslip motion relative to cell organelle motion, even at microns away from the coverslip, we microfabricated 10-µm-high and 3-µm-diameter cylindrical SU-8 posts on glass coverslips to serve as fiduciary markers during tracking. Microposts were fabricated using SU-8 2010, a negative epoxy-type photoresist (MicroChem, Newton, MA, USA), on 40-mm circular no.1 glass coverslips (Bioptechs, Butler, PA, USA) using standard photolithography techniques (7). Glass coverslips were initially cleaned using, sequentially, acetone for 10 min, isopropanol for 10 min, and deionized (DI) water for 10 min. An adhesion layer (OmniCoat™; MicroChem) was spin-coated on the glass surface using a spin coater (PWM32; Headway Research, Garland, TX, USA) followed by a baking process on a hot plate at 200°C for 1 min. SU-8 2010 was then spin-coated on top of the adhesion layer at 5000 rpm for 30 s to obtain a 10-µm-thick uniform photoresist layer. The resulting layer was baked at 65°C for 10 min and at 95°C for 20 min.

A chromium mask was designed in AutoCAD® (Autodesk, San Rafael, CA, USA) and fabricated commercially (MEMS and Nanotechnology Exchange, Reston, VA, USA) to the desired specifications. The mask and glass samples were exposed under ultraviolet (UV) light for 30 s to selectively activate the SU-8 2010 layer and baked at 50°C for 15 min and 95°C for 20 min. Samples were then developed using SU-8 Developer (MicroChem) to remove inactivated SU-8 2010 and adhesion layers and washed in isopropanol solution. Finally, the coverslips were baked at 150°C for 3 min (Figure 1A). Dimensions of posts were verified with sequential z-axis optical sectioning using DIC microscopy and a piezo-controlled stage Nano-View™/ M (Madcity Labs, Madison, WI, USA). Additional verification was performed using a stylus (Alpha-Step 200; KLA-Tencor, San Jose, CA, USA).

Figure 1. Fabrication and imaging of fiduciary microposts. (Click to enlarge)

Bovine aortic endothelial cells (BAECs; VEC Technologies, Rensselaer, NY, USA) were subcultured between passages 3–10 on the processed coverslips (Figure 1, B and C). MCDB-131 complete medium supplemented with 11% fetal bovine serum (FBS), 10 ng/mL epidermal growth factor (EGF), 1 µg/mL hydrocortisone, 100 mg/500 mL EndoGro® (VEC Technologies), 45 mg/500 mL heparin, and 5 mL/500 mL 100× antibiotic/antimycotic solution was used as the culture medium. Cells were incubated at 37°C with 5% CO2, 90% humidity. In order to maintain the physiological environment while imaging, a closed system flow chamber, [Focht Chamber System (FCS2®); Bioptechs] was used. The FCS2 flow chamber was mounted on a Nano-View computer-controlled motorized microscope stage (Mad City Labs, Madison, WI, USA).

Cells were imaged under DIC microscopy using an Olympus IX71 inverted research microscope (Olympus America, Center Valley, PA, USA) with a 60× PlanApo oil immersion objective, numerical aperture (NA) 1.45. To further improve the resolution in the DIC mode, a 60× LUMPlanFL water immersion objective (NA 0.90; Olympus America) was used in place of the condenser. An immersion oil with refractive index 1.33 (Series AAA; Cargille Laboratories, Cedar Grove, NJ, USA) was used in place of water to prevent evaporation of immersion medium. Images were acquired in real-time with a high-resolution, 12-bit SensiCam QE CCD camera (The Cooke Corporation, Romulus, MI, USA). Camware 2.12 software (The Cooke Corporation) was used to acquire time-lapse, 8-bit TIFF images at 15–30 frames/s with exposure times of 1–3 ms with an image resolution of 0.1 µm/pixel. Multiple particle tracking of microposts and endogenous cell organelles was performed using an image correlation algorithm with centroid calculation-based subpixel interpolation method written in Lab VIEW™ (National Instruments, Austin, TX, USA). Briefly, the centroid of the maximum correlation of a template image of the feature being tracked and its subsequent images in a time series were plotted.

Tracking of post and vesicle images verified readily distinguishable low- and high-frequency coverglass motion artifacts. When the motorized microscope stage was on, it exhibited a preferential drift in the Y direction with an average velocity of 0.3 µm/s (Figure 2A). The presence of microposts as fiduciary markers in the timelapse images enabled subtraction of the low frequency directional drift of the microscope stage from the cell organelles motion to reveal their natural trajectories (Figure 2B).

Figure 2. Particle tracking of micropost and vesicle. (Click to enlarge)

Particle tracking measurements are subject to fluctuations in light source intensity, camera electronics, and shot noise of the CCD chip. Since the microposts and vesicles were tracked simultaneously, subtraction of micropost position corrected for the errors due to system vibration and shot noise. To quantify such indeterminate errors (8), precision of our tracking algorithm was calculated as the standard deviation of the square root of the mean squared displacement (MSD) of a micropost and was found to be 2.4 nm.

Time-dependent position autocorrelation function (MSD) of endogenous vesicles was computed before and after subtracting stage drift to evaluate how well the procedure reduced system noise over a frequency range from 0.1 to 30 Hz (Figure 2C). This technique facilitated the segregation of vesicle trajectories into diffusive, subdiffusive, and directed motion components and thus provided the ability to measure the Brownian dynamics as well as motordriven active transport of molecules in cells. Importantly, tracking analysis was carried out on vesicles that were about 2–5 µm away from the coverslip. Markers confined to the coverslip would not be useful for subtracting out the global stage and coverslip motion.

Conventional analog as well as motorized, piezo- and galvanometer-driven microscope stages exhibit temperature-related drifts over long time periods (minutes to hours) (9) even when motors are turned off. This low-frequency stage drift can be quantified by tracking the stationary microposts and can be easily subtracted out from the particle trajectories during post-processing. It is anticipated that three-dimensional (3-D) fiduciary markers could also serve to correct for long-term changes in focus by assessing the characteristic 3-D refraction patterns. Similarly, image processing of fiduciary microposts accompanied by feedback control of stage position could facilitate longrange scanning applications, where it is important to return to precisely the same location. Such methods could improve high-throughput screening in scanning cytometry experiments (10).

In conclusion, the technique of microfabricating fiduciary markers on glass coverslips along with multiple particle tracking algorithms provides a cost-effective solution to eliminate low- and high-frequency microscope stage and coverglass motion from time-lapse images of adherent cells. Potentially, microcontact printing can be used to add fluorescent tracer molecules to the microposts for corrections in coverglass motion under confocal and fluorescence imaging (11,12), making fiduciary microposts applicable for the tracking of single molecules in cells.