A Multicolor 3D-STORM High-Resolution Visualization Protocol for Collagen Mineralization in Self-Assembled Recombinant Type I Collagen Fibrils

Summary

Here, we present a protocol to distinguish intrafibrillar versus extrafibrillar mineralization in recombinant collagen fibrils using multicolor 3D‑STORM, integrating optimized labeling, imaging, and quantitative colocalization analysis.

Abstract

This protocol describes a multicolor three-dimensional stochastic optical reconstruction microscopy (3D-STORM) method for nanoscale visualization of collagen mineralization in a recombinant type I collagen self-assembled fibril model. The method enables simultaneous imaging of collagen, non-collagenous proteins (e.g., chondroitin sulfate), and calcium phosphate mineral phases. Sample preparation involves amino‑silanization and collagen self‑assembly, followed by mineralization using a calcium phosphate medium that forms amorphous calcium phosphate (ACP) at an early stage (30 min) and matures into hydroxyapatite (HAP) by 6 h. Multiplexed immunofluorescence labeling is then performed, and samples are first assessed by confocal microscopy before 3D-STORM image acquisition using an oxygen-scavenging imaging buffer. Data processing and analysis are carried out using publicly available software. Compared to conventional electron or confocal microscopy, this protocol combines molecular specificity with nanoscale resolution (typical lateral precision 20–30 nm, axial 50–60 nm), allowing three‑dimensional visualization of intrafibrillar versus extrafibrillar mineralization patterns. Representative results show clear visualization of collagen networks, associated non-collagenous proteins, and mineral phases within three-dimensional space. Quantitative metrics including Pearson’s correlation coefficient (0.89 ± 0.04) and Manders’ overlap coefficient (0.91 ± 0.03) are provided in the Results section. This protocol offers a powerful tool for researchers in biomaterials science, biomineralization, and bone tissue engineering who require nanoscale insight into mineralization dynamics.

Introduction

Collagen mineralization is a fundamental biological process pivotal in the formation of hard tissues such as bones and teeth¹. The intricate structure of collagen fibers, coupled with finely tuned regulation of mineral deposition, endows remarkable mechanical strength and structural integrity to these tissues². Collagen serves not merely as a passive scaffold but as an active participant, orchestrating precise mineral deposition through complex molecular and physical interactions³. Elucidating these mechanisms is crucial for understanding pathological conditions such as osteoporosis and dental caries, and for developing biomimetic materials for regenerative therapies⁴.

The diameter of collagen fibers in hard tissues ranges from approximately 50 to 100 nm, with hydroxyapatite (HAP) particles being even smaller (typically 2–5 nm in thickness and 20–30 nm in length) and intercalated within fibrillar gaps⁵. While gap zones can act as nucleation sites, in native tissues, minerals initially form in interfibrillar spaces and subsequently expand into intrafibrillar compartments. Traditional characterization methods include histological staining, confocal laser scanning microscopy^6,7, and electron microscopy^8,9. Histological staining provides a macroscopic assessment but cannot evaluate nanoscale mineralization states. Confocal microscopy enables observation of specific components but is diffraction-limited (~200 nm laterally), unable to resolve intrafibrillar versus extrafibrillar mineralization¹⁰. Electron microscopy offers high resolution but lacks molecular specificity. Although immunogold labeling can provide molecular specificity for electron microscopy, it requires specialized processing and is less amenable to multiplexed, three-dimensional visualization of multiple components compared to STORM, and involves lengthy experimental cycles and high costs¹¹.

Stochastic optical reconstruction microscopy (STORM) overcomes the diffraction limit by localizing individual fluorophores with high precision, achieving ~20 nm lateral resolution¹². In comparison to other super-resolution techniques such as stimulated emission depletion (STED)¹³microscopy and structured illumination microscopy (SIM)¹⁴, STORM offers higher localization precision (~20 nm laterally and ~50 nm axially) and is compatible with a wider range of organic fluorophores. In particular, STED requires specialized dyes and high laser powers that can damage biological samples, while SIM offers only ~100 nm resolution, insufficient to resolve fibril‑scale (50–100 nm) features. STORM provides a practical balance of resolution, multiplexing capability, and sample compatibility. Published guidelines have described standardized test samples to facilitate optimization of STORM imaging parameters and resolution assessment¹⁵. When combined with three-dimensional imaging capabilities, 3D-STORM enables nanoscale visualization of multiple components simultaneously¹⁶. Recent advances have extended STORM to multicolor and multiplexed imaging, enabling visualization of multiple targets within the same sample^17,18.

This protocol uses a biomimetic in vitro model based on self-assembled recombinant type I collagen fibrils and is explicitly designed for in vitro recombinant type I collagen self‑assembled fibril models to distinguish intrafibrillar from extrafibrillar mineralization at the nanoscale. It is not suitable for live‑cell imaging, as STORM requires fixed samples and oxygen‑scavenging buffers. It is also not suitable for native highly mineralized tissues (e.g., mature bone) without prior decalcification or antigen retrieval. If only overall mineral density or large‑area morphology needs assessment, conventional confocal or electron microscopy is more efficient.

The overall goal of this protocol is to provide a standardized, stepwise workflow for using multicolor 3D-STORM to visualize the nanoscale distribution of minerals and non-collagenous proteins within individual collagen fibrils, with specific emphasis on distinguishing intrafibrillar versus extrafibrillar mineralization. This protocol integrates optimized immunolabeling with a tailored imaging buffer to enable simultaneous tracking of multiple organic components (collagen, non-collagenous proteins) and inorganic phases (ACP, HAP) in three dimensions¹⁹. A key innovation is the quantitative assessment of intrafibrillar versus extrafibrillar mineralization patterns. Quantification is achieved by two independent criteria: (1) calculating Pearson’s colocalization coefficient between the mineral (a calcium indicator dye (e.g., calcein)) and collagen (a far‑red fluorescent dye) channels using the Colocalization module in image analysis software (coefficient >0.8 indicates strong association); (2) analyzing the persistence of mineral signal throughout Z‑slices of individual fibrils: if mineral signal is present in central slices (Z = 0 to ±120 nm from the fibril’s vertical center) at intensity ≥50% of the maximum, it is classified as intrafibrillar. Unlike conventional electron or confocal microscopy, this protocol combines molecular specificity with nanoscale resolution, enabling three-dimensional spatial distribution of intrafibrillar mineralization.

This protocol is designed for researchers in biomaterials science, biomineralization, and bone tissue engineering requiring nanoscale insight into mineralization dynamics. The standardized procedure can also be adapted to study other mineralized collagen systems, such as demineralized dentin slices or collagen-based hydrogels, by adjusting initial sample preparation steps accordingly.

Protocol

All experiments involving biological samples were conducted in accordance with the guidelines and regulations of the Core Facilities, Zhejiang University School of Medicine and were approved by the Institutional Biosafety Committee (Approval Certificate No. BSL20235710079). The experimental protocol described herein utilizes commercially sourced reagents and in vitro biomimetic systems. It does not involve human participants, animal subjects, or human tissue samples, and therefore does not require ethical approval from an institutional review board.

CAUTION: All procedures involving hazardous chemicals must be performed in a fume hood with appropriate personal protective equipment (lab coat, gloves, safety goggles). Dispose of chemical waste according to institutional regulations. For NaN₃, collect waste in a dedicated container marked “azide waste” and do not mix with acids (risk of explosive gas). For glutaraldehyde, inactivate with 10% excess sodium bisulfite before disposal. For β‑mercaptoethanol, oxidize with bleach (1:10 v/v) for 1 h before drain disposal.

NOTE: Immunofluorescence labeling MUST be performed BEFORE mineralization to avoid epitope masking by mineral deposits. For applications requiring post‑mineralization labeling, antigen retrieval may be needed.

1. Preparation of mineralization medium

1. Preparation of mineralization medium for reconstituted collagen fibrils
   1. Prepare calcium stock solution by adding 100 µL of 5 mg/mL polyaspartic acid (p-Asp, Mw 9-11 kDa) dropwise to 5 mL of 3.44 mM CaCl₂ solution in a 15 mL conical tube.
   2. Vortex mix in the tube for 30 s until homogeneous.
   3. Add p‑Asp slowly and mix thoroughly to stabilize amorphous calcium phosphate (ACP).
   4. Add 5 mL of phosphate solution (19 mM Na₂HPO₄ + 300 mM NaCl) to the calcium stock solution.
   5. Vortex mix for 1 min.
      NOTE: Final concentrations after mixing: 1.67 mM CaCl₂, 9.5 mM Na₂HPO₄, 150 mM NaCl, and 50 µg/mL p-Asp.
   6. Prepare the mineralization medium immediately before use. Do not store; use the unstable ACP precursor immediately to achieve consistent results.
      NOTE: Storage leads to premature crystallization.
2. Preparation of mineralization medium for collagen gels/demineralized dentin
   1. Prepare calcium-containing solution by dissolving 8.77 g NaCl, 0.01 g NaN₃, 6.06 g Tris, and 1.11 g CaCl₂ in 800 mL deionized water.
   2. Bring the volume to 1 L with deionized water.
   3. Add 350 µg/mL polyacrylic acid (PAA, Mw ~1800) to the calcium-containing solution.
   4. Vortex for 1 min.
      NOTE: Use PAA instead of p‑Asp for better stabilization at higher calcium concentrations.
   5. Prepare phosphate solution by dissolving 0.85 g Na₂HPO₄ in 1 L deionized water to obtain 6 mM Na₂HPO₄.
   6. Filter-sterilize the phosphate solution using a 0.22 µm membrane.
   7. Combine equal volumes (e.g., 500 mL each) of calcium and phosphate solutions with magnetic stirring at 300 rpm.
   8. Adjust pH to 7.4 using 1 M HCl or NaOH while monitoring with a pH meter.
      NOTE: Maintain pH at 7.4 ± 0.1. Do not allow pH deviations >0.2, which cause premature crystallization.

2. Preparation of recombinant collagen fibers

1. Amino-silanization of glass-bottom dishes
   NOTE: This treatment uses (3-aminopropyl) triethoxysilane (APTES) to introduce positively charged amino groups (‑NH₂) onto the glass surface, which promotes electrostatic adsorption and stabilization of negatively charged collagen monomers.
   1. Add 200 µL of APTES solution (5% v/v in absolute ethanol) to each glass-bottom culture dish (35 mm, #1.5H thickness, 0.17 mm ± 0.005 mm).
   2. Ensure complete coverage of the dish surface with the APTES solution.
   3. Incubate in darkness for 2 h at room temperature (20–25 °C).
   4. Rinse the dishes three times with absolute ethanol (2 mL per wash).
   5. Ultrasonicate in deionized water for 10 min at 40 kHz. Remove residual APTES completely to prevent nonspecific binding.
   6. Silanized dishes can be stored dry at 4 °C for up to 1 week.
   7. (Optional) Bake for enhanced stability: Place washed and dried dishes in an oven at 100–120 °C for 30–60 min.
   8. Allow to cool to room temperature before proceeding.
      NOTE: If using plastic dishes, reduce temperature to below 80 °C to prevent deformation. Adapt the silanization procedure from Bitter and Muir²⁰.
2. Collagen self-assembly and crosslinking
   1. Pipette 100 µL of collagen-I solution (50 µg/mL in 0.1 M acetic acid) onto each silanized dish.
   2. Incubate at 37 °C for 10 h in a humidity chamber (> 90% relative humidity). Maintain stable humidity to prevent solution evaporation.
   3. Verify fibril formation with phase‑contrast or confocal microscopy; a successful preparation shows a fine, web‑like fibril network.
   4. To verify proper fibril formation at the ultrastructural level, prepare a separate sample on a polyvinyl formal/carbon-coated TEM grid following the same self-assembly conditions (collagen-I 50 µg/mL, 37 °C, 10 h, humidity > 90%).
   5. Stain with 1% phosphotungstic acid (pH 7.0) for 10 min.
   6. Wash with deionized water.
   7. Examine under a transmission electron microscope. Successful fibrillogenesis is indicated by the characteristic 67 nm D-periodic banding pattern (see Figure 3).
   8. Carefully aspirate residual collagen solution from the glass-bottom dishes.
   9. Add 2 mL of chondroitin sulfate (CS) solution (50 mg/mL in PBS, pH 5.5) to each dish.
   10. Incubate at 4 °C for 12 h to allow electrostatic adsorption of CS onto collagen fibrils.
   11. Prepare EDC/NHS crosslinking solution²¹: dissolve 0.2 M EDC (1-ethyl-3-(3-dimethylaminopropyl) carbodiimide) and 0.05 M NHS (N-hydroxysuccinimide) in MES buffer (0.1 M MES, pH 5.5).
       NOTE: MES buffer can be prepared by dissolving MES free acid in deionized water and adjusting pH to 5.5 with NaOH.
   12. Remove the CS solution and add 2 mL of the EDC/NHS crosslinking solution.
   13. Incubate at room temperature for 2 h to covalently immobilize CS on collagen.
   14. Wash dishes three times with PBS (5 mL each, 5 min each) to remove unreacted reagents.
   15. Store processed collagen fibers in PBS at 4 °C for up to 24 h before mineralization.

3. Immunofluorescence labeling (performed BEFORE mineralization)

1. Sample preparation
   1. Rinse collagen fibrils three times with 500 mM NaCl solution (10 mL per wash, 5 min each).
   2. Wash three times with deionized water (10 mL per wash, 5 min each).
   3. Perform washes on a horizontal shaker platform rotating at 50 rpm to ensure thorough washing while minimizing shear forces that could detach assembled fibrils.
2. Blocking and primary antibody incubation
   1. Incubate with blocking buffer (5% bovine serum albumin in PBS) at 37 °C for 1 h in a humidified chamber.
   2. Prepare antibody cocktail in blocking buffer: rabbit anti-collagen-I (1:100 dilution) and mouse anti-chondroitin sulfate (1:100 dilution).
   3. Add 200 µL of antibody cocktail per sample.
   4. Cover samples with laboratory sealing film.
   5. Incubate at 4 °C overnight (12–16 h).
   6. Protect from light using aluminum foil. After primary antibody incubation, samples can be stored in PBS at 4 °C for up to 24 h.
3. Secondary antibody staining
   1. Remove unbound primary antibodies by washing the dishes three times with washing buffer (0.1% Tween-20 in PBS), 10 min each wash.
   2. Prepare the fluorescent secondary antibody mixture in blocking buffer (200 µL per 35 mm dish) containing: goat anti-rabbit IgG conjugated to a far‑red fluorescent dye (1:100) and goat anti-mouse IgM conjugated to a red fluorescent dye (1:100).
      NOTE: From this step onward, protect samples from light to prevent photobleaching of the fluorophore.
   3. Incubate samples at room temperature (20–25 °C) for 1 h in the dark.
   4. Wash dishes three times with washing buffer (10 min each) to remove unbound secondary antibodies.
      NOTE: Store labeled samples in PBS at 4 °C (light‑protected) for up to 48 h before imaging. Include no-primary-antibody controls to check for autofluorescence.

4. Mineralization of collagen fibers (performed AFTER labeling)

1. Mineralization process
   1. Add 2 mL of freshly prepared mineralization medium per dish.
   2. Incubate at 37 °C for 30 min (short-term) for early-stage mineral (ACP) formation.
   3. Alternatively, incubate at 37 °C for 6 h (long-term) for mature mineral (HAP) crystallization.
      NOTE: Maintain exact temperature at 37 °C ± 0.5 °C using a calibrated incubator.
   4. Gently aspirate mineralization medium.
   5. Rinse three times with deionized water (5 mL per wash, 1 min intervals).
2. Calcium phosphate labeling
   1. Add 4 µM of a calcium indicator dye (e.g., calcein) in PBS (200 µL per dish).
   2. Incubate at room temperature (20–25 °C) for 30 min in the dark.
      NOTE: The calcium indicator dye (e.g., calcein) binds to calcium ions and does not discriminate between amorphous and crystalline phases; phase assignment is based on mineralization time (30 min = ACP, 6 h = HAP). It should be protected from light using amber tubes or foil wrap.
   3. Rinse five times: three washes with deionized water and two washes with 70% ethanol (1 min each), alternating between the two.
3. Sample preparation for imaging
   1. Immerse samples in imaging buffer: 10% (w/v) glycerol in PBS. Optionally add an antifade reagent according to the manufacturer's instructions.
   2. Apply a sufficient volume (20–50 µL) of imaging buffer to cover the sample.
   3. Place a coverslip over the sample.
   4. Store samples at 4 °C in the dark for up to 72 h.

5. Observation under a laser confocal microscope

1. Transfer of samples from 4 °C storage to room temperature (20–25 °C).
2. Allow 10 min equilibration to prevent condensation.
3. Maintain light protection using amber containers or foil wrap.
4. Verify that the imaging buffer covers the entire sample.
5. Configure the confocal microscope with the following settings:
   1. For collagen imaging (far‑red fluorescent dye): 647 nm laser, emission filter 650-710 nm.
   2. For calcium phosphate imaging (calcium indicator dye (e.g., calcein)): 488 nm laser, emission filter 500-550 nm.
   3. Objective: 100× oil immersion (NA 1.4).
   4. Pinhole diameter: 1 Airy Unit (AU).
   5. Pixel dwell time: 1-2 µs.
      ​NOTE: Perform laser alignment and pinhole calibration before imaging.
6. Perform sequential scanning: first scan with 647 nm excitation (collagen).
7. Perform second scan with 488 nm excitation (mineral).
   NOTE: Collect channels sequentially to avoid bleed-through.
8. For 3D reconstruction, set Z-stack step size to 0.5 µm or smaller to ensure adequate sampling.
9. Save files with metadata preserved.
10. For colocalization analysis, use image analysis software capable of calculating Pearson's correlation coefficient (e.g., publicly available software with suitable plugins).

6. Imaging by Three-dimensional stochastic optical reconstruction microscopy (3D-STORM)

1. Preparation of STORM imaging buffer²²
   NOTE: Glucose oxidase and catalase are added to the imaging buffer to scavenge oxygen, thereby reducing photobleaching and prolonging the blinking of fluorophores, which is essential for single-molecule localization in STORM.
   1. Prepare a glucose oxidase (GOx) stock solution at 8 mg/mL in 50 mM sodium acetate buffer (pH 5.1).
   2. Prepare a catalase stock solution at 160 µg/mL in 1× PBS.
      NOTE: Aliquot the enzyme stocks, snap-freeze them using liquid nitrogen or a dry-ice/ethanol bath, and store at -20 °C or -80 °C. Avoid repeated freeze-thaw cycles.
   3. Prepare fresh STORM imaging buffer on ice, protected from light.
   4. Combine the following components in the order listed to a final volume of 1 mL: sterile nuclease-free H₂O (to 1 mL), 1 M NaCl (10 µL, 10 mM final), 1 M Tris-HCl pH 8.0 (50 µL, 50 mM final), freshly prepared 1 M cysteamine (MEA) adjusted to pH ~7.0 (50 µL, 50 mM final), 50% (w/v) glucose (200 µL, 10% w/v final), GOx stock (200 µL, 1.6 mg/mL final), catalase stock (200 µL, 32 µg/mL final).
   5. Mix gently by pipetting or inversion. Do not vortex.
      NOTE: Buffer must be prepared fresh and kept on ice protected from light. Use within 30 min of preparation. For detailed optimization of STORM imaging conditions, refer to established protocols using test samples¹⁵.
   6. Add 200 µL of freshly prepared STORM imaging buffer to fully cover the sample.
2. 3D-STORM system configuration
   1. Ensure the imaging path is directed to an electron‑multiplying CCD (EMCCD) camera.
   2. Engage the cylindrical lens module for 3D imaging.
   3. Set software to 3D STORM acquisition mode.
   4. Configure optical path filters for the dyes used.
   5. Turn on the 647 nm laser and set the power to a low level (1-5 mW).
   6. Using a 100× oil immersion objective (NA ≥1.4), locate the region of interest in widefield or TIRF live preview mode.
   7. Acquire a conventional wide-field fluorescence image for later comparison.
   8. Switch to TIRF illumination mode.
   9. Calibrate the TIRF angle to achieve a thin illumination field (~100-200 nm) to minimize background fluorescence.
   10. Adjust camera exposure time (10-30 ms) based on initial signal intensity.
   11. Adjust imaging laser power based on initial signal intensity.
   12. Perform a brief test acquisition.
   13. Verify that parameters are set correctly without causing rapid photobleaching.
       NOTE: Ensure precise alignment of the cylindrical lens axis with the sample plane. Most modern systems have automated calibration. For manual calibration, use 100 nm fluorescent beads to verify symmetric, round point spread functions (PSFs).
   14. Begin with 10–30 ms exposure at a 647 nm laser power of 1–2 kW/cm².
   15. Verify that the blinking density is 0.1–1 molecule/µm².
   16. If the density is too low or too high, adjust the laser power or exposure time accordingly.
   17. Repeat the verification and adjustment until the ideal density is achieved.
3. 3D-STORM data acquisition
   1. Set acquisition software to "Continuous Activation" mode.
   2. Set imaging laser (647 nm) to high power (100-500 mW fiber input) to transition most fluorophores to the dark state.
   3. Set activation laser (405 nm) to low power (0.1–5 mW).
   4. Optimize 405 nm power empirically to maintain 100–200 localized molecules per frame in a 256×256-pixel region.
   5. Set total number of frames to 10,000–20,000.
   6. Use 1 frame of activation light (405 nm) followed by 5 frames of imaging light (647 nm) per cycle.
   7. Start acquisition.
   8. Operate EMCCD with maximum sensitivity (EM gain applied)
   9. Use 10–30 ms exposure per frame.
   10. During acquisition, monitor active molecule density.
   11. Adjust 405 nm laser power to maintain a steady stream of single-molecule events.
       ​NOTE: Raw movie data can be stored on a standard hard drive for long-term archiving before reconstruction.
4. Data analysis of collagen mineralization (using image analysis software with STORM module)
   1. In the analysis software, select the appropriate camera driver (STORM module).
   2. Click [Analysis GUI] in the STORM panel. The analysis window opens.
   3. Click [File Open]. Select the raw microscopy image file to be analyzed. Click Open.
   4. Determine the minimum fluorescence value (Minimum Height) for valid signals.
   5. Find the darkest spot still recognizable as a single‑molecule signal.
   6. Move the mouse to its center.
   7. Read the Peak Height value.
   8. Record this value.
   9. Click [Identification Settings]. In the dialog, set the following parameters: Minimum Height (enter the value recorded in step 6.4.8), Maximum Height (20000), CCD Baseline (100), and for 3D‑STORM data check “3D” (uncheck for 2D).
   10. Click >> to expand more parameters. Set: Minimum Width (nm) to 200, Maximum Width (nm) to 700 (400 for 2D), Initial Fit Width (nm) to 300, Max Axial Ratio to 2.5 (1.3 for 2D), and Max Displacement (pix) to 1.
   11. Click OK to close the dialog.
   12. Perform a test analysis to validate the parameters. Uncheck Drift Correction.
   13. Set Periods to 1.
   14. Click [Test].
   15. After analysis, click OK in the pop‑up dialog.
   16. Check that signal spots are correctly identified. Background noise should not be mis‑selected. True spots should not be missed.
   17. If unsatisfactory, repeat steps 6.4.9–6.4.16 to adjust parameters until correct.
   18. After the test analysis passes, perform the full analysis. Check Drift Correction.
       NOTE: The analysis software performs drift correction using a redundant cross‑correlation (RCC) algorithm that maximizes correlation between time segments of molecular localizations. No fiducial beads are required²³.
   19. Keep Periods = 1.
   20. Click [Start] (or click [Test] again, then [Start]).
   21. Wait for the analysis to be completed. Click OK in the pop‑up dialog.
   22. After analysis, two result files (.bin and .txt) are generated in the same folder as the.nd2 file.
   23. For colocalization analysis (647 nm and 488 nm channels), display both channels simultaneously in the channel list of the analysis window.
   24. Select an appropriate region of interest (ROI).
   25. Use the colocalization module to calculate Pearson's correlation coefficient. If not automatically provided, manually export point cloud data to a colocalization plugin in publicly available image analysis software. Record the value.
   26. Perform Z‑slice analysis to confirm intrafibrillar mineralization. In the main menu, select View > Z‑stepping > Slice.
   27. Extract individual planes at 60 nm intervals.
   28. Observe the mineral signal (green, 488 nm channel). Check whether the signal persists in the central slices (Z = 0 nm ±120 nm). Persistence confirms intrafibrillar mineralization.
   29. Optional: perform density filtering to reduce over‑counting from densely packed emitters. Open the STORM reconstructed image in publicly available image analysis software using a plugin for single‑molecule localization analysis²⁴.
   30. To reduce over‑counting from densely packed emitters, apply density filtering (e.g., median filter or local density threshold) based on the signal density.
   31. If drift correction was not enabled in step 6.4.18 or additional correction is needed, perform drift correction. Correct based on the PSF of fiducial markers. Alternatively, use cross‑correlation algorithms (e.g., drift correction implemented in the same plugin).
   32. Perform spectral unmixing to eliminate channel crosstalk. In the STORM analysis window, click the crosstalk removal tool (usually in the lower right corner).
   33. Set the radius to 25.0 nm (recommended). Select the source channel. Select the removal method: Statistical is recommended.
   34. To remove cross‑activation caused by the excitation laser, select Subtract in addition to NSA > Cross Activation. Click OK. A new channel, Nonspecific Activation‑Xt, is automatically generated.
   35. Validate autofluorescence levels and background signals using unstained control samples. Ensure all signals are labeled specifically.
   36. Perform 3D visualization. Select a region of interest (ROI) in the analysis window.
   37. Click the [3D] button (or [Show 3D]). Generate a 3D projection or volume rendering.
   38. Right‑click to adjust rendering parameters. Export videos in common formats (e.g., .avi or .mp4).
   39. To classify a mineral signal as intrafibrillar, perform Z‑slice analysis as in 6.4.26–6.4.28. Export intensity profiles.
   40. Define the fibril center as the Z‑plane where the collagen signal is maximal.
   41. A mineral signal is considered intrafibrillar if its intensity at Z = 0 (center) and at Z = ±60 nm is ≥50% of the maximum mineral intensity in that fibril. If the signal drops below 30% at Z = 0, classify as extrafibrillar.
   42. Record the classification for at least 50 fibrils per condition to calculate the percentage intrafibrillar mineralization.
       NOTE: Alternative publicly available plugins are also available for density filtering, drift correction, and spectral unmixing.

Results

Successful implementation of this protocol yields a high‑resolution three‑dimensional visualization of mineralized collagen fibrils using multicolor 3D‑STORM. The following results illustrate typical outcomes, quality controls, and quantitative assessments.

Figure 1 shows a multicolor 3D‑STORM reconstruction of a collagen network mineralized with amorphous calcium phosphate (ACP). Collagen (labeled with a far‑red fluorescent dye) appears as a well‑defined fibrillar network. Chondroitin sulfate (GAG, labeled with a red dye) is closely associated with the collagen fibrils, and regions of colocalization appear cyan in the merged image. Importantly, ACP particles (green, labeled with a calcium indicator dye) are observed within the boundaries of collagen fibrils, demonstrating intrafibrillar mineralization at the early stage (30 min). This figure highlights the protocol’s ability to simultaneously resolve three distinct components (collagen, GAG, mineral) at the nanoscale, a feat not achievable with conventional confocal microscopy (see Figure 5 for comparison of imaging resolution). The nanoscale colocalization of ACP within fibrils confirms that the amorphous precursor can infiltrate the fibrillar interior before crystalline transformation.

Figure 2 provides quantitative evidence for intrafibrillar penetration of mature hydroxyapatite (HAP) after 6 h of mineralization. Figure 2A presents 2D STORM images showing extensive colocalization of collagen (red) and HAP (green), with merged channels appearing yellow. Figure 2B is a 3D volume reconstruction illustrating integrated architecture. Figure 2C shows Z‑axis slice analysis at 60 nm intervals from the top (Z = −120 nm) to the center (Z = 0) and to deeper sections (Z = +120 nm). The HAP signal persists in the central slices (Z = 0 to ±120 nm) at an intensity ≥50% of the maximum, which meets the classification criteria for intrafibrillar mineralization defined in Protocol steps 6.4.39–6.4.41. Quantitative colocalization analysis from three independent experiments (n = 3 replicates, each with 5 regions of interest) yielded a Pearson’s correlation coefficient of 0.89 ± 0.04 and a Manders’ overlap coefficient of 0.91 ± 0.03 (mean ± SD). These values indicate a strong and specific association between collagen and HAP within the fibrils, confirming that the mature crystalline phase also resides intrafibrillarly.

Figure 3 validates the structural integrity of the self‑assembled collagen scaffold using transmission electron microscopy. Collagen fibrils stained with 1% phosphotungstic acid (pH 7.0) display the characteristic 67 nm D‑periodic cross‑banding pattern, which is diagnostic of native‑like fibrillogenesis. This quality control step is essential before proceeding to mineralization experiments, as it confirms that the scaffold is structurally intact and capable of supporting intrafibrillar mineral deposition. Without this banding pattern, the collagen may be denatured or improperly assembled, leading to artifactual mineralization patterns.

Figure 4 illustrates a representative negative result obtained when mineralization conditions are not properly controlled (e.g., the absence of polyaspartic acid or pH > 7.6). Under these suboptimal conditions, HAP deposits (green) are observed exclusively on the glass substrate outside the collagen fibrils (red), with no intrafibrillar invasion. This outcome serves as an important control: it demonstrates that the intrafibrillar mineralization observed in Figures 1 and 2 is not due to nonspecific precipitation or incomplete washing but rather requires precise control of the chemical environment (pH 7.4 ± 0.1, presence of polyaspartic acid, and immediate use of fresh mineralization medium). Researchers should include such negative controls to validate their own system.

Figure 5 shows representative confocal microscopy images acquired before STORM acquisition for preliminary sample screening. The collagen channel (Figure 5A) reveals a clear fibrillar network, and the HAP channel (Figure 5B) shows mineral associated with the fibrils. The merged image (Figure 5C) confirms colocalization at the diffraction‑limited level (~200 nm). These images serve two purposes: (1) they verify sample quality (adequate labeling, minimal aggregation, and specific mineral association) before proceeding to time‑consuming STORM imaging; and (2) they guide selection of regions of interest for 3D‑STORM acquisition. Importantly, the confocal images lack the resolution to distinguish intrafibrillar versus extrafibrillar mineralization. Note that the mineral signal appears continuous along the fibrils without revealing whether it is inside or outside. This limitation underscores the need for the super‑resolution approach described here.

Figure 6 presents two control experiments. Figure 6A (no‑primary‑antibody control) shows no specific signal when only the secondary antibody is applied, confirming that the observed signals in labeled samples are not due to nonspecific secondary antibody binding. Figure 6B (unstained control) shows no fluorescence signal (completely black), ruling out significant autofluorescence from the sample or substrate. These controls are essential for validating the specificity of the immunofluorescence labeling. Any detectable signal in these controls would indicate the need to adjust blocking or washing steps.

Under our optimized imaging conditions, the 3D‑STORM system achieved typical localization precision of 20–30 nm laterally and 50–60 nm axially, consistent with the original 3D‑STORM literature²⁵. Drift correction was performed during post‑processing using the built‑in redundant cross‑correlation (RCC) algorithm, which eliminates the need for exogenous fiducial markers²³. For phase assignment, ACP was assigned at 30 min mineralization and HAP at 6 h, based on the established maturation kinetics. The ability to distinguish these two phases temporally, combined with nanoscale localization, allows researchers to study the dynamics of intrafibrillar mineral transformation.

In summary, the representative results demonstrate that this protocol enables nanoscale distinction between intrafibrillar and extrafibrillar mineralization in a recombinant collagen model, with quantitative colocalization metrics and appropriate negative controls. The method is particularly valuable for researchers studying biomineralization mechanisms, biomimetic materials, and bone tissue engineering.

Supplementary Table 1 summarizes common problems encountered during the mineralization and STORM imaging procedures, along with their possible causes and recommended solutions. Please click here to download this file.

Figure 1: Multicolor 3D-STORM reconstruction of mineralized collagen fibrils. Collagen (red, far‑red fluorescent dye) and chondroitin sulfate (GAG, blue, red fluorescent dye) are imaged simultaneously. Colocalization of collagen and GAG appears as magenta in the merged channel, and regions where all three overlaps appear white. Amorphous calcium phosphate (ACP, green, calcium indicator dye) is also shown. ACP is observed within the boundaries of collagen fibrils, indicating intrafibrillar localization. Scale bar: 1 µm. Please click here to view a larger version of this figure.

Figure 2: 3D-STORM imaging of intrafibrillar hydroxyapatite (HAP) mineralization. (A) 2D STORM images of collagen (red, far‑red fluorescent dye), HAP (green, calcium indicator dye), and merged channel. (B) 3D volume reconstruction. (C) Z-axis slice analysis at 60 nm intervals (Z = −120, −60, 0, +60, +120 nm). Persistent HAP signal in the central slices (Z = 0 to ±120 nm) meets the classification criteria for intrafibrillar mineralization (intensity ≥50% of maximum). Quantitative colocalization calculated from this figure: Pearson's r = 0.89 ± 0.04, Mander's overlap = 0.91 ± 0.03. Scale bars: 0.1 µm. Please click here to view a larger version of this figure.

Figure 3: Transmission electron microscopy (TEM) validation of self-assembled collagen fibrils. Collagen fibrils were stained with 1% phosphotungstic acid (pH 7.0). The diagnostic 67 nm D-periodic cross-banding pattern confirms successful fibrillogenesis and structural integrity. Scale bar: 200 nm. Please click here to view a larger version of this figure.

Figure 4: Representative negative result: extrafibrillar mineralization. Collagen (red, far‑red fluorescent dye) and HAP (green, calcium indicator dye). HAP deposits exclusively on the glass substrate outside collagen fibrils, with no intrafibrillar invasion. This suboptimal outcome is included as a negative control to illustrate the range of possible results. Scale bar: 0.1 µm. Please click here to view a larger version of this figure.

Figure 5: Representative confocal images used for preliminary screening. (A) The collagen (red, far‑red fluorescent dye) channel shows a clear fibrillar network. (B) HAP (green, calcium indicator dye) channel shows mineral association. (C) The merged image shows colocalization of collagen and HAP. Scale bar = 2 µm. Please click here to view a larger version of this figure.

Figure 6: Control experiments. (A) No‑primary‑antibody control: only secondary antibody applied; no specific signal. (B) Unstained control: no fluorophore or antibody applied; no fluorescence signal (completely black). Scale bar = 1 µm. Please click here to view a larger version of this figure.

Supplementary Table 1: Troubleshooting guide. Common problems encountered during mineralization and 3D-STORM acquisition/analysis, along with their possible causes and recommended solutions. See text for detailed step numbers.Please click here to download this file.

Discussion

This protocol provides a comprehensive workflow for nanoscale visualization of collagen mineralization using multicolor 3D-STORM. Several critical steps require particular attention to ensure successful outcomes.

First, sample preparation is foundational for high-quality STORM imaging. The amino-silanization of glass-bottom dishes must be thorough to ensure stable attachment of collagen fibrils throughout subsequent washing and labeling steps. Residual APTES can cause nonspecific binding and high background, while incomplete silanization may lead to fibril detachment. The recommended ultrasonication step (10 min at 40 kHz) effectively removes unbound silane while preserving surface functionalization. For crosslinking collagen fibers, EDC/NHS is recommended²¹ for high specificity. Alternatively, 0.1% glutaraldehyde can be used (incubate 30 min at room temperature, then wash thoroughly), but it is less specific. Crucially, we recommend verifying proper fibril formation by TEM before proceeding to mineralization. As shown in Figure 3, the presence of the characteristic 67 nm D-periodic banding pattern confirms that the self-assembly process has produced structurally intact, native-like collagen fibrils²⁶. This quality control step prevents misinterpretation of results arising from poorly formed or denatured collagen scaffolds.

Second, the mineralization medium must be prepared with precise pH control (7.4 ± 0.1) and used immediately. The amorphous calcium phosphate (ACP) precursor is highly sensitive to pH and ionic strength; small deviations can cause premature crystallization. Therefore, the mineralization medium must be used immediately after preparation. For studies requiring comparisons across multiple time points, prepare fresh medium for each experiment rather than using stored solutions. When mineralization conditions are not properly controlled (e.g., insufficient polyaspartic acid or incorrect pH), extrafibrillar mineralization predominates, as shown in  Figure 4. In this negative control, HAP deposits exclusively on the glass substrate outside collagen fibrils, with no intrafibrillar invasion. The inclusion of such negative results is essential to validate that the intrafibrillar signal observed in Figure 1 and Figure 2 is indeed due to controlled intrafibrillar mineralization and not to nonspecific precipitation or incomplete washing. Furthermore, previous studies have emphasized that distinguishing intrafibrillar from extrafibrillar mineralization requires careful control of the local chemical environment and the presence of stabilizing additives such as polyaspartic acid²⁷.

Third, immunofluorescence labeling requires careful optimization. The antibody concentration (1:100) provided works well for the collagen and chondroitin sulfate system described but may need titration for different antibodies or sample types. Always include no-primary-antibody controls to assess autofluorescence and non-specific binding. From the secondary antibody step onward, strict light protection is essential to prevent fluorophore photobleaching.

Fourth, the STORM imaging buffer must be prepared fresh and used within 30 min. The oxygen-scavenging system (glucose oxidase/catalase) loses activity over time, and cysteamine is both light-sensitive and oxygen-sensitive. Pre-aliquoting enzyme stocks and storing them at -80 °C ensures consistent performance across experiments. The blinking density should be monitored in real time and adjusted by modulating the 405 nm laser power; too few molecules prolong acquisition time, while too many cause overlapping PSFs and reduced localization precision. For detailed optimization of STORM imaging parameters, we refer readers to established test sample protocols¹⁵.

Fifth, data processing requires standardized parameters for meaningful comparisons across samples. The minimum photon count threshold (typically 500-1000 photons) excludes low-confidence localizations. If sample signals are weak, it may be appropriately reduced to 300 photons, but it is not recommended to go below 200 photons. Drift correction using fiducial markers or cross-correlation algorithms is essential for maintaining resolution, particularly for 3D reconstructions²⁸. Spectral unmixing helps eliminate crosstalk between channels, which is critical for accurate colocalization analysis. Troubleshooting common problems is outlined in Supplementary Table 1.

The protocol has several limitations. It is optimized for biomimetic in vitro models; application to native, highly mineralized tissues (e.g., mature bone) may require additional steps such as decalcification or more aggressive antigen retrieval, which could affect ultrastructure. Reliance on specific antibodies may introduce labeling density issues or steric hindrance, particularly in densely packed structures. The technique is equipment-intensive, requiring access to a high-end STORM microscope with appropriate laser lines and an EMCCD camera. Additionally, the total time required (~53 h from sample preparation to data analysis) may limit throughput for some applications.

Despite these limitations, this protocol offers significant advantages over alternative methods. Compared to electron microscopy, it provides molecular specificity through immunofluorescence labeling, enabling simultaneous visualization of multiple organic and inorganic components. Compared to confocal microscopy, it achieves ~10-fold higher spatial resolution, enabling distinction between intrafibrillar and extrafibrillar mineralization patterns. The 3D capability provides volumetric information essential for understanding mineral distribution within the collagen matrix.

The method has broad applicability in biomineralization research. Potential applications include studying the role of non-collagenous proteins in mineral nucleation, evaluating biomimetic materials for bone regeneration, investigating pathological mineralization in diseases such as osteoporosis and dental caries, and assessing the effects of therapeutic interventions on mineral distribution. With appropriate modifications, the protocol can be adapted to study other organic-inorganic interfaces in tissues or biomaterials.

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
Polyaspartic acid (p-Asp)	Sigma-Aldrich	P9903	Stabilizer for amorphous calcium phosphate
Calcium chloride (CaCl2)	Sigma-Aldrich	C1016	Calcium source
Sodium phosphate dibasic (Na2HPO4)	Sigma-Aldrich	S0876	Phosphate source
Sodium chloride (NaCl)	Sigma-Aldrich	S9888	Ionic strength adjuster
Polyacrylic acid (PAA)	Sigma-Aldrich	323667	Stabilizer for high-concentration calcium
Tris base	Sigma-Aldrich	T1503	Buffer component
Sodium azide (NaN3)	Sigma-Aldrich	S2002	Antimicrobial agent
(3-Aminopropyl)triethoxysilane (APTES)	Sigma-Aldrich	440140	Glass surface functionalization agent
Absolute ethanol	Sigma-Aldrich	459836	Solvent
Type I collagen solution (50 μg/mL in 0.1 M acetic acid)	Corning	354249	Self-assembly scaffold
Chondroitin sulfate (CS)	Sigma-Aldrich	C9819	Non-collagenous protein mimic
EDC (1-ethyl-3-(3-dimethylaminopropyl)carbodiimide)	Sigma-Aldrich	E7750	Crosslinker
NHS (N-hydroxysuccinimide)	Sigma-Aldrich	130672	Crosslinker activator
MES free acid	Sigma-Aldrich	M5287	Buffer for crosslinking
Phosphate-buffered saline (PBS)	Gibco	10010023	Washing and dilution buffer
Bovine serum albumin (BSA)	Sigma-Aldrich	A3059	Blocking agent
Rabbit anti-collagen-I antibody	Abcam	ab34710	Primary antibody for collagen
Mouse anti-chondroitin sulfate antibody	Sigma-Aldrich	C8035	Primary antibody for CS
Goat anti-rabbit IgG conjugated to far-red fluorescent dye (Alexa Fluor 647)	Thermo Fisher Scientific	A-21244	Secondary antibody for collagen
Goat anti-mouse IgM conjugated to red fluorescent dye (Alexa Fluor 568)	Thermo Fisher Scientific	A-11031	Secondary antibody for CS
Calcein (calcium indicator dye)	Sigma-Aldrich	C0875	Calcium phosphate label
Tween-20	Sigma-Aldrich	P1379	Detergent for washing buffer
Glycerol	Sigma-Aldrich	G5516	Imaging buffer component
Glucose oxidase (GOx)	Sigma-Aldrich	G7141	Oxygen scavenger
Catalase	Sigma-Aldrich	C1345	Oxygen scavenger
Cysteamine (MEA)	Sigma-Aldrich	M6500	Thiol for fluorophore blinking
D-Glucose	Sigma-Aldrich	G6152	Substrate for glucose oxidase
Sodium acetate	Sigma-Aldrich	S2889	Buffer for GOx stock
Hydrochloric acid (HCl)	Sigma-Aldrich	320331	pH adjustment
Sodium hydroxide (NaOH)	Sigma-Aldrich	71690	pH adjustment
Phosphotungstic acid	Sigma-Aldrich	P4006	Negative stain for TEM
Glass-bottom culture dishes (35 mm, #1.5H)	MatTek	P35G-1.5-14-C	Sample substrate; thickness 0.17 mm
Ultrasonic cleaner (40 kHz)	Branson	B200	Cleaning device
Humidity chamber	Thermo Fisher Scientific	11-432-10	For collagen self-assembly
Transmission electron microscope	Hitachi	HT7800	TEM imaging
Formvar/carbon-coated TEM grids (200 mesh)	Sigma-Aldrich	FCF200-Cu	TEM sample support
Horizontal shaker platform	Labnet	S2030-RC	Gentle washing
Confocal laser scanning microscope	Nikon	A1	Preliminary screening
3D-STORM microscope system (with 405/488/647 nm lasers, cylindrical lens, EMCCD)	Nikon	N-STORM	Super-resolution imaging
100× oil immersion objective (NA 1.49)	Nikon	MRD01991	High-resolution imaging
pH meter	Mettler Toledo	FiveGo F2	pH control
STORM acquisition and analysis software	Nikon	NIS-Elements (STORM module)	STORM data acquisition and processing
.nd2 file format (raw microscopy image file)	Nikon	N/A	Raw image file format generated by Nikon microscopes.
Publicly available image analysis software	Open source	N/A	e.g., ImageJ with ThunderSTORM plugin for single-molecule localization analysis (colocalization, drift correction)
Parafilm	Bemis	PM996	Sample covering during incubation
Aluminum foil	Any laboratory supplier	N/A	For light protection (e.g., wrapping samples)
Amber microcentrifuge tubes	Fisher Scientific	05-669-21	For light protection of fluorophores
Coverslips (No. 1.5)	Corning	2855-18	Sample mounting