Background

Fluorescence microscopy has become a cornerstone technique in biological research, enabling visualization of molecular structures and dynamic processes within cells and tissues. However, its optical resolution is fundamentally limited to approximately 250 nm laterally and 500 axially [1,2], leaving many subcellular structures unsolved. Although advanced super-resolution methods such as STED, SIM, STORM, and PALM have been developed to overcome this limitation, they often require specialized instrumentation, complex sample preparation, and intensive computational processing [1,3–5].

Expansion microscopy (ExM) provides a cost-effective and accessible alternative by physically enlarging biological specimens embedded in a swellable polyacrylamide-based hydrogel [5–7]. ExM improves the effective resolution in proportion to the expansion factor, enabling nanoscale visualization of cellular and tissue architecture using standard fluorescence microscopes [6,8–10]. Despite its simplicity and versatility, ExM imaging remains challenged by sample drift, primarily caused by subtle movements of the soft, water-rich hydrogel during three-dimensional (3D) image acquisition. Such drift introduces spatial distortions that degrade image quality and compromise the accuracy of quantitative analyses, particularly in large or highly expanded samples [7,11–13].

A variety of experimental approaches, such as poly-L-lysine coating, agarose embedding, and mechanical stabilization, have been explored to minimize drift. However, these methods often yield inconsistent results or introduce undesirable artifacts, including sample shrinkage or fluorescence loss [11–14]. Consequently, post-acquisition drift correction has emerged as a more reliable and noninvasive strategy to ensure accurate 3D reconstruction and quantitative measurement.

Existing software tools, including several ImageJ plugins [15,16], provide limited functionality for drift or distortion correction in ExM datasets. Most were originally designed for time-lapse image registration rather than z-stack alignment, and thus often require metadata modification, manual parameter tuning, or correlation-based alignment methods that lack the precision needed for quantitative ExM analysis.

To address this gap, we developed 3D-Aligner, a dedicated computational tool optimized for drift correction in 3D fluorescence microscopy datasets, and validated its performance, particularly using ExM images [14]. 3D-Aligner is fully integrated into our previously developed 3D-Speckler platform, designed for nanometer-scale particle analysis in fluorescence microscopy [2,17]. The 3D-Aligner software detects faint, nonspecific background fluorescence signals and applies a nearest-neighbor matching algorithm between adjacent z-planes to determine drift trajectories with nanometer precision. By relying on background features rather than biological structures, 3D-Aligner minimizes correction-induced artifacts and ensures unbiased alignment.

The tool offers a streamlined, user-friendly interface with customizable parameters for feature size, expected drift range, and channel selection. Users can visualize drift trajectories and corrected image stacks to facilitate downstream quantitative analysis. Benchmarking results demonstrate that the 3D-Aligner outperforms existing registration programs in both correction accuracy and reconstruction fidelity across diverse expansion factors and labeling conditions [14].

Beyond ExM, the 3D-Aligner can be applied to particle tracking, motion analysis, or distortion correction in other fluorescence microscope methods. We also demonstrated its capability to perform nanoscale particle tracking in live-cell imaging datasets. Overall, the 3D-Aligner provides a robust, accessible, and precise solution for drift correction, significantly enhancing the reliability of quantitative 3D imaging in expansion microscopy and related super-resolution methods. In this protocol, we present a detailed, step-by-step guide for configuring and using 3D-Aligner, illustrated with example drifted datasets from 4-fold expanded samples using the 4×3D-ExM method, which represents one of the most commonly used ExM approaches in a wide range of research fields [6].