Western Blot Loading Controls: GAPDH, Actin, or Total Protein?

If you've ever probed a western blot for GAPDH or beta-actin as a loading control, you're in good company — it's been the default approach for decades. But the field has moved. Major journals now recommend total protein normalization over single housekeeping proteins, and reviewers are increasingly asking why you chose the control you did.

Here's what changed, why it matters, and what to do about it.

The Problem with Housekeeping Proteins

The assumption behind using GAPDH, beta-actin, or tubulin as a loading control is simple: these proteins are expressed at constant levels regardless of experimental condition. The problem is that assumption is often wrong.

• Expression varies with treatment. Aldridge et al. (2008, J Neurosci Methods) showed that GAPDH and beta-actin can change more than 2-fold under common experimental conditions — drug treatments, serum starvation, hypoxia. If your loading control changes with your treatment, your normalization is meaningless.
• Housekeeping proteins saturate at low loads. Taylor & Posch (2014, Biomed Res Int) demonstrated that housekeeping proteins like GAPDH typically saturate above ~4 µg of total protein per lane. Above that, signal plateaus even as you load more protein. Since most experiments load 10–30 µg/lane, the "loading control" may be flat across all lanes regardless of actual loading differences — exactly the opposite of what you want.
• Single-protein controls are noisy. Janes (2015, Sci Signal) showed that normalization to a single protein produces a coefficient of variation (CV) around 21%. Averaging across three or more loading controls drops that CV to ~7%.

The Total Protein Alternative

Total protein normalization (TPN) measures the total amount of protein in each lane rather than a single protein. Common methods include:

• Ponceau S staining — reversible red stain applied to the membrane before antibody probing. Cheap, simple, widely available.
• Stain-free technology — Bio-Rad and other vendors offer gels with a trihalo compound that makes proteins fluorescent under UV after a brief activation step. No staining required.
• REVERT total protein stain — LI-COR's fluorescent total protein stain. Compatible with near-infrared imaging systems.
• Coomassie or SYPRO Ruby — post-transfer staining of the membrane or gel.

Eaton et al. (2013, PLOS One) showed that total protein normalization produces significantly more reproducible results than single housekeeping protein controls. Taylor & Posch (2014) demonstrated that total protein signal is linear from ~10–80 µg/lane — far wider than the ~4 µg saturation point of housekeeping proteins.

What the Journals Say

The shift isn't just academic opinion. It's in the author guidelines:

• JBC (Journal of Biological Chemistry): Explicitly recommends total protein normalization. If you use a housekeeping protein, you must justify the choice and demonstrate it doesn't change under your experimental conditions (JBC Author Guidelines, 2024).
• Cell Press: Requires demonstration that any housekeeping protein used as a loading control is unchanged by the experimental treatment.
• Nature journals: Recommend total protein normalization and require clear reporting of the normalization strategy used.

When Single-Protein Controls Are Still OK

Total protein normalization isn't always practical, and single-protein controls can be valid when used carefully:

• You've validated the control for your specific conditions. If you can show (with data) that your housekeeping protein doesn't change across your treatment groups, it's a defensible choice.
• You're working with a tissue where TPN is unreliable. Some tissue types (e.g., heavily glycosylated or lipid-rich samples) don't stain evenly with Ponceau or stain-free methods.
• You load well below the saturation point. If you're loading 1–2 µg total protein per lane, housekeeping protein saturation is less of a concern.
• You use multiple housekeeping proteins. Janes (2015) showed that averaging three or more loading controls substantially improves normalization accuracy, even when each individual control is imperfect.

Phospho-Protein Normalization: A Special Case

If you're quantifying a phosphorylated protein (e.g., p-ERK, p-AKT), don't normalize to GAPDH or total protein. Normalize to the total (unphosphorylated + phosphorylated) target protein instead. The ratio you want is phospho-signal / total-target-signal, with both background-subtracted and both within the linear range. The JBC and LI-COR guidelines are explicit about this.

Practical Recommendations

1. Default to total protein normalization (Ponceau, stain-free, REVERT) for new experiments. It's more reproducible, has wider dynamic range, and is what reviewers expect.
2. If you use a housekeeping protein, validate it. Show that it doesn't change across your treatment conditions. Include this data in your supplementary figures.
3. Don't let your loading control saturate. If you're loading 20 µg/lane and using GAPDH, there's a good chance it's saturated. Run a loading titration to check.
4. For phospho-proteins, normalize to total target protein. Not to a housekeeping protein. Not to total protein.
5. Report your choice. Whatever method you use, state it clearly in your methods section. Reviewers shouldn't have to guess.

Tools like VoilaBlot support both housekeeping protein and total protein normalization workflows. Whichever approach fits your experiment, the normalization math, background subtraction, and QC checks happen automatically — so you can focus on the biology rather than the spreadsheet.

References

• Aldridge GM, Podrebarac DM, Greenough WT, Weiler IJ. (2008) J Neurosci Methods 172(2):250-254.
• Eaton SL et al. (2013) PLOS One 8(8):e72457.
• Gilda JE, Gomes AV. (2013) Anal Biochem 440(2):186-188.
• Janes KA. (2015) Sci Signal 8(371):rs2.
• JBC Author Guidelines. (2024) ASBMB JBC Resources.
• Taylor SC, Posch A. (2014) Biomed Res Int 2014:361590.