A step-by-step protocol for the scratch assay begins with seeding cells to 90-95% confluence, starving them in low-serum medium for 12-24 hours, then creating a uniform wound with a sterile 200 µL pipette tip held perpendicular to a straight-edge guide. Wash twice with warm PBS to remove debris, add low-serum medium (0.5-2% FBS) with mitomycin C at 10 µg/mL if you need to isolate migration from proliferation, and image the same fixed coordinates at 0, 6, 12, and 24 hours. Measure wound width or area with ImageJ's polygon or line tool, normalize to the 0-hour field, and report closure percentage plus the migration rate in µm/hour across at least three biological replicates.
What the scratch assay actually measures
The scratch assay, also called the wound-healing assay, is an in vitro migration model. You grow a confluent monolayer, mechanically disrupt it, and track how cells move into the cell-free gap over time. It is cheap, uses standard culture equipment, and yields kinetic data on two-dimensional collective migration.
The mechanism is contact-inhibition release. In a confluent monolayer, cells sense neighbors through cadherin-based adherens junctions and stop moving freely. When you scratch, cells at the wound edge lose those contacts on one side and gain polarization cues. Actin polymerizes into a leading-edge lamellipodium, integrins engage fibronectin and collagen deposited on the plastic, and the sheet advances by combined protrusion, contraction, and rear retraction.
In plain terms: the assay does not measure one cell running into a gap. It measures a coordinated sheet front, where leader cells pull and follower cells fill in.
Two distinct processes drive closure, and this distinction governs everything downstream. Proliferation adds new cells and can close a wound without migration. Migration redistributes existing cells. A scratch that closes in serum-rich medium reflects both. Per the widely cited 2012 report by Liang, Park, and Guan in Nature Protocols, this confounding is the assay's central weakness and must be controlled explicitly.
The step-by-step protocol in numbered order
Each step below is a decision point, not just a handling instruction. Where a number matters, the working range is given.
Step 1. Seed cells at 5 × 10^4 to 2 × 10^5 cells per well in a 12- or 24-well plate, or 2-5 × 10^5 per well in a 6-well plate. Target 90-95% confluence in 24-48 hours. Underconfluent monolayers give ragged, non-reproducible edges.
Step 2. Serum-starve for 12-24 hours before scratching. This synchronizes the cell cycle and lowers baseline mitotic drive so that early closure reflects migration. Skip this and your 0-6 hour data is contaminated by S-phase entry.
Step 3. Scratch with a sterile 200 µL pipette tip, held perpendicular to a straight-edge ruler laid along the plate bottom. Apply consistent pressure to reach the plastic. A wobbling tip produces a wound wider at one end, breaking measurement comparability.
Step 4. Wash twice with warm PBS or serum-free medium to remove lifted debris. Debris re-adheres and is counted as closure by some segmentation tools, inflating your rate by 10-20%.
Step 5. Add starvation medium (0.5-2% FBS) with or without mitomycin C at 10 µg/mL pre-treatment for 1-2 hours. Mitomycin C crosslinks DNA and blocks proliferation, isolating migration. Note its tradeoff in the table below.
Step 6. Image at 0 hours at fixed stage coordinates, then at 6, 12, and 24 hours. Mark the plate underside with a grid or use a motorized stage to return to identical fields. Field drift is the single most common source of false closure.
Step 7. Analyze in ImageJ or a comparable tool. Draw the wound boundary at each time point, compute area, and normalize to 0 hours. Report percent closure and µm/hour.
Timeline: how fast a scratch should close
Closure speed depends on cell type, substrate coating, and medium. The table below reflects typical ranges for common lines under low-serum migration conditions.
| Time point | Expected state | Typical wound width remaining | Action |
|---|---|---|---|
| 0 hours | Clean, debris-free wound | 500-1000 µm (200 µL tip) | Image reference field |
| 2 hours | Cell spreading, no net closure | 95-100% of initial width | Check edge uniformity |
| 6 hours | Lamellipodia polarized, closure starts | 70-90% of initial width | First rate estimate |
| 12 hours | Active collective migration | 40-70% of initial width | Midpoint for dose-response |
| 18 hours | Sheet nearing mid-wound | 20-45% of initial width | Watch edge collisions |
| 24 hours | Approaching or reaching confluence | 0-30% of initial width | Endpoint for most lines |
| 48 hours | Fully closed or over-confluent | 0% (indistinguishable from fill-in) | Too late; data uninterpretable |
A well-proliferating line such as HeLa can close a 500 µm wound in under 24 hours in 10% FBS. Slow lines such as primary keratinocytes may need 48-72 hours in low serum. Endpoint choice must match cell cycle time: if your doubling time is 20 hours, closure at 24 hours is partly proliferative unless blocked.
Migration versus proliferation: controls that decide the result
The comparison below summarizes the standard ways to separate migration from proliferation, a distinction competitors often skip.
| Approach | What it blocks | Data quality | Caveat |
|---|---|---|---|
| Low serum (0.5-2% FBS) | Reduces, does not eliminate division | Good for fast lines | Slow lines arrest entirely |
| Mitomycin C (10 µg/mL, 1-2 h) | DNA synthesis | Strong migration isolation | Cytotoxic at high doses; can reduce motility |
| Serum starvation 24 h | Cell-cycle synchrony | Improves reproducibility | Stress responses may alter phenotype |
| Proliferation index (EdU/BrdU) | Measures division alongside closure | Best quantitative control | Adds assay complexity |
| Cytochalasin D (0.1-1 µM) | Actin polymerization | Positive migration-negative control | Collapses all motility; not a probe |
| Time-lapse imaging | Distinguishes individual from collective | Gold standard | Requires live-cell microscope |
The non-obvious point: even with mitomycin C, closure is not pure migration. Treated cells still spread and change shape, and some lines respond to the drug with reduced lamellipodial dynamics. The most defensible design combines low serum, mitomycin C, and an independent proliferation readout in a parallel plate.
Edge cases and failure modes that break reproducibility
Edge irregularity is the most common hidden variable. A 200 µL tip creates a wound with a ragged but statistically consistent edge after compression; two operators rarely produce identical wounds. Use one tip type, one ruler, and one operator per experiment.
Cell density drift between wells changes closure rate by roughly 15-25% for a 10% difference in starting confluence, because contact inhibition strength scales with density. Plate all wells from one suspension.
Evaporation at the plate edge concentrates medium in outer wells. Never use perimeter wells for quantification, or use a plate shaker-free humidified tray.
Bubbles and floating debris inflate segmentation. If your analysis software auto-thresholds, debris reads as foreground. The wash step matters more than it appears.
Treatment compounds that alter proliferation will mask or mimic antimigratory effects. Always run the proliferation control for any drug study.
Correcting three widespread misconceptions
Misconception one: a scratch assay proves cell migration. It proves wound closure, which can be driven by spreading, proliferation, or contraction of the sheet. Only time-lapse tracking of individual cells establishes migration.
Misconception two: faster closure means more aggressive cells. In cancer biology, this is often assumed. Yet the direction of effect can reverse: some genes increase both migration and proliferation while also increasing contact inhibition, slowing net closure. Closure rate is a composite of opposing forces.
Misconception three: 24 hours is a universal endpoint. Late closure conflates proliferation, cell division, and density-dependent slowing. Per the 2012 Nature Protocols guidance, endpoints should fall in the linear phase of the closure curve, roughly before the wound reaches 30% of its original width.
What to actually do before, during, and after each run
Positive actions that improve signal, listed because the protocol is only as good as its controls.
- Pilot the closure curve first, imaging every 2 hours up to 48 hours to find your line's linear phase and set endpoints there.
- Fix cell number per well from a single-counted suspension and verify confluence visually at 90-95% before starving.
- Mark a coordinate grid on the plate underside and image the identical field each time; record stage positions in the imaging software.
- Run at least three biological replicates on separate days and three technical fields per replicate.
- Include a mitomycin C control and a proliferation index in every comparison arm.
- Report closure as both percent of initial area and µm/hour to make the rate independent of wound size.
Limits of the model and when to choose another assay
The scratch assay is two-dimensional, uses an artificial wound, and lacks the extracellular matrix architecture of real tissue. It cannot distinguish directed chemotaxis from random motility without a gradient chamber. It cannot capture single-cell invasion through matrix.
For invasion, use a Transwell or Matrigel assay. For single-cell tracking, use time-lapse with live-cell microscopy. For gradient responses, use a microfluidic chemotaxis device. The scratch assay is the fastest, cheapest screen; it is not the definitive mechanism test.
One practical caveat from professional reviews: because wound edges experience different mechanical tension than interior cells, closure rates can differ 20-40% between a scratch made with a 200 µL tip and a uniform silicone insert. If absolute rates matter, standardize the wound geometry with an insert rather than a pipette tip.