Antibodies 101: Normalization and Loading Controls for Western Blots

By Emily P. Bentley

Ah, the notorious western blot: we meet again. So useful, yet so finicky to design and optimize. Today we’ll cover the normalization and loading controls needed for relative quantification of a western blot — and why you might want to be careful relying on so called “housekeeping proteins.”

Fully quantifying a western blot requires a standard curve with a pure sample of your protein of interest (POI). This approach is reliable, but it’s overkill in many cases. Often, you may not need to know the exact quantity of your POI, but simply whether its concentration changes in response to experimental parameters. This is a job for a semi-quantitative western blot.

Most western blots benefit from a positive control — a loading control — that you know is present in every lane of your gel. At their most basic, loading controls help you rule out errors like uneven pipetting or irregular membrane transfer.

Cartoon representations of western blots with multiple lanes are shown. The first blot does not have a loading control. A band is present in lanes 1, 2, and 4, but not lane 3. The blot is captioned, “Did condition 3 eliminate my protein, or did I skip a lane by accident?” A forked arrow leads from this panel to two possible outcomes with a loading control (blots 2 and 3). Blot 2 shows the same band pattern as blot one, labeled protein of interest (POI), as well as a loading control band present in all four conditions. This blot is captioned, “Definitely a real effect!” Blot 3 shows the same band pattern for the POI, but condition 3 lacks a loading control band as well as a POI band. This blot is captioned, “Whoops!”
Figure 1: A loading control is a type of positive control necessary to demonstrate a POI’s absence. It also helps confirm that lanes are loaded evenly. Created with BioRender.com.

Loading controls are even more important for semi-quantitative western blots. No pipetter is perfect, and proteins may not transfer precisely evenly across the entire blot. So how can you accurately compare different lanes? In relative quantification, the POI bands are normalized to the loading control: the intensity of each POI band is divided by the intensity of the loading control in the same lane, and the resulting fraction is reported.

Cartoon representations of two western blots are shown along with graphical representations of mathematical normalization. Bands can be described as light, medium, or dark gray. Both western blots show a POI band and a loading control band for two conditions.   On both blots, condition 1 has a light gray POI band, and condition 2 has a medium gray POI band.   On the left blot, the loading control band is dark gray for both conditions. The normalized values are represented graphically as fractions, with light gray divided by dark gray for condition 1 producing a smaller value than medium gray divided by dark gray for condition 2.  On the right blot, the loading control band is medium gray for condition 1 and dark gray for condition 1. The normalized values are represented graphically as fractions, with light gray divided by medium gray for condition 1 producing an equal value to medium gray divided by dark gray for condition 2.
Figure 2: Normalization with loading controls. Under each lane, the fraction of the POI band over the loading control is shown. In the left gel, the loading controls are equally strong, so the normalized value simply reports which POI band is stronger. The right gel has exactly the same POI bands as the left gel, but because the loading control band is weaker in condition 1 than condition 2, the normalized values for the two conditions are equal. Created with BioRender.com.

This kind of normalization is a nearly ubiquitous tool, but proceed with caution! Loading controls only address one type of error that can impact your western blot, and depending too much on normalization can amplify other errors in experimental design and gel quantification.

A flow chart representing values measured with a western blot, errors that influence those values, and controls to correct for those errors.  The top row of the chart shows each of the following items connected to the next by a solid arrow to represent cause and effect: 1.	Protein concentration in sample 2.	Protein abundance on membrane 3.	Band intensity  The final item leads to a cartoon of Blugene holding a western blot. From this cartoon, dotted arrows lead backward along the same row of items, representing how researchers use band intensity (item #3) to calculate protein abundance on membrane (item #2) and infer protein concentration in sample (item #1).  The second row shows three sources of error. Each item has an arrow leading to one of the steps in the first row, merging with the solid arrows to show the contribution of error to the result. Below each source of error is a control, connected to the error it is used to manage with a flat inhibition arrow.  Error #1: Unexpected experimental perturbations. This error influences the protein concentration in sample and is managed with experimental controls.  Error #2: Loading and transfer errors. This error influences the protein abundance on membrane and is managed with loading controls.  Error #3: Quantification errors. This error influences band intensity and is managed with linear detection tests.
Figure 3: In a perfect world, protein concentration in your sample would directly correlate to protein abundance on your membrane, which would be directly reported by band intensity. Researchers use later steps in this chain (blue) to infer or calculate things about earlier ones. In practice, multiple sources of error (red) influence these relationships and must be addressed with separate controls (green). Created with BioRender.com.

Researchers often rely on assumptions that their loading controls are independent of these other sources of error. It’s important to validate these assumptions for your particular experiment and choice of loading control. First, let’s explore why these assumptions are key to using loading controls accurately.

Assumption #1: The loading control is not perturbed by the experiment

The loading control is supposed to control for, well, loading errors. When using it to normalize your results, you are assuming that all differences in the loading control are due to loading and transfer errors — not to any factors from before the proteins even hit the gel.

But what if your experiment changes the expression of your loading control without your knowledge? If you can’t trust Assumption #1, any change in the normalized intensity of your POI could reflect either a genuine change in your POI or an opposite change in your loading control.

Assumption #2: The loading control can be accurately quantified


In western blotting, the relationship between protein abundance and band intensity is an S-shaped curve. Sparse proteins may not be detected at all, while overloaded proteins may saturate the detector. In between these extremes lies the linear detection range.

A cartoon graph with “protein abundance” on the x axis and “band intensity” on the y axis. The relationship between the two is shown with a sigmoidal line, and the linear portion of this line is indicated.
Figure 4: The linear detection range of the assay is between the two dotted lines. Created with BioRender.com.

To see why this S-shaped relationship poses problems for proteins of very different abundance, consider the following example gel.

A cartoon western blot showing two conditions, each with a POI band and a loading control band. Each of these four bands is also represented on a cartoon plot of protein abundance vs band intensity as a colored dot plotted on the sigmoidal line shown in Figure 4.  In condition 1, the POI band is light gray, and the loading control is dark gray. In condition 2, the POI band is medium gray, and the loading control is very dark gray. The colors of two loading control bands are difficult to distinguish by eye, although they are different on close inspection.  This is reflected in the placement of each band on the sigmoidal plot. The POI bands fall within the linear detection range. Condition 2 has a higher protein abundance and a higher band intensity than condition 1. However, the loading control bands fall above the linear detection range, where the sigmoidal curve flattens out. Although the condition 2 band has a higher protein abundance, it has an almost equal band intensity to condition 1.
Figure 5: An example gel (left) with bands corresponding to points on the plot of protein abundance versus band intensity (right). Created with BioRender.com.

On this gel, the protein of interest band is significantly lighter in “condition 1” than “condition 2.” By eye, the loading control appears about the same between the two conditions. But it isn’t! (No really, I promise it isn’t.) The intuitive, qualitative interpretation of this gel is misleading.

We can see why on the S-shaped curve. On the protein abundance axis, both blue “condition 1” points (POI and loading control) are smaller than the analogous green “condition 2” points by the exact same value. For the two solid POI points, this difference is reflected on the band intensity axis. But the loading control points are outside of the linear detection range. As a result, they have almost the same value on the band intensity axis.

This effect causes misleading qualitative conclusions, and it’s also problematic for computational quantification. Outside the linear detection range, band intensity measurements become extremely susceptible to error, since tiny changes in intensity can reflect massive differences in protein abundance.

Validating the assumptions for different loading controls

So how can you address these errors without making incorrect assumptions? The answer depends on the loading control you choose.

Housekeeping proteins as loading controls

Ubiquitously expressed proteins like β-actin or GAPDH are popular choices for loading controls. These are often called “housekeeping proteins” and are detected with an antibodies, just like the POI.

This approach is so common that we’ve used it for our cartoon gels so far. However, it throws both assumptions discussed above into question.

Assumption #1: The loading control is not perturbed by the experiment

Unfortunately, multiple reports suggest that common housekeeping proteins are influenced by a variety of experimental conditions (Ghosh et al., 2014). Although this assumption is likely true in many cases, you should validate it for your particular setup.

We recommend starting by searching the literature to see what is already known about your housekeeping protein in your context. If you don’t find anything, you’ll need to conduct a control experiment to demonstrate that expression of your reference protein does not vary under your experimental conditions. Otherwise, you might want to consider a different loading control.

Assumption #2: The loading control can be accurately quantified

Most housekeeping proteins are highly expressed in cells. This means that at sample concentrations appropriate to detect a less-abundant POI, the loading control will almost certainly be overloaded. One paper called this problem “the most common error associated with Western blotting quantification” (Gilda & Gomes, 2013).

To address this source of error, it’s important to determine the linear detection range of all proteins that will be measured in your assay. This is influenced by sample concentration as well as by antibody dilution — both factors we’ve covered in our previous blog posts (1, 2) on western blotting. If the abundance of your POI and loading control aren’t too far apart, they may have overlapping linear detection ranges, allowing you to accurately quantify both on the same blot.  

Total protein measurements as loading controls

Total protein measurements are an accurate alternative loading control option. Total protein stains or stain-free imaging systems are both effective for measuring the total protein present in a lane (Aldridge et al., 2008; Gilda & Gomes, 2013). Depending on the approach you choose, your loading control measurement will take place either before or after blotting for your POI.

Two cartoon western blots. One shows a POI blot with two conditions and different band intensities. The other shows a total protein measurement, with the entire lane filled with proteins of different size and abundance. The total protein measurement is nearly identical between the two conditions.
Figure 6: Total protein measurements are taken on the same gel or membrane (depending on your approach) as your final blot. Quantifying the entire lane of protein provides a loading control that can be used for normalization. Created with BioRender.com.


Assumption #1: The loading control is not perturbed by the experiment

By relying on the total protein in the lane, your loading control won’t be strongly influenced by unexpected variations in any individual protein. This ensures you are truly comparing the abundance of your POI as a fraction of the total sample.

Assumption #2: The loading control can be accurately quantified

Because the total protein is measured in a separate step from the POI blot, you don’t need to worry about the linear detection ranges of multiple targets. Convenient!

Comparison of loading control approaches

Cartoons are nice and all, but we’ll leave you with this real-world comparison of different loading control approaches from Gilda & Gomes, 2013.

Three panels of data are shown.  Panel A shows a western blot photograph of liver lysate probed with β-actin. Each lane shows a higher total concentration of protein, from 10-45 µg in 5 µg increments. There is not a clear gradient of darker bands at higher concentrations of lysate.  Panel B shows a plot of the relative intensity quantification of panel A. The relative intensity ranges from 1.0 at 10 µg total protein to approximately 2.0 at 45 µg total protein, and does not increase linearly. There are 8 total points, and thus 7 possible comparisons of adjacent points; in 3 out of 7 comparisons, the second point’s intensity is lower than or similar to that of the preceding point. The error bars on each point vary widely.  Panel C shows a plot of the relative intensity quantification of two different total protein measurements, Ponceau S and Stain free. (The total protein blots are not shown here, but can be found in the cited paper.) The intensity of both measurements starts at 1.0 at 10 µg total protein. At 45 µg total protein, the Ponceau S measurement has a relative intensity of approximately 4.0, while the stain-free measurement has a relative intensity of approximately 2.5. Both measurements increase linearly and monotonically, with similar error bars on each point.
Figure 7: Semi-quantification of liver lysate using different loading controls. A) Western bot of liver lysate (1040 µg) probed with β-actin (1:2000). Representative of three independent blots. BC) Graphs showing the relative intensity of the β-actin (B) or total protein (C) on the membrane versus the amount of protein loaded on the gel (n=3). Data is presented as mean ± SEM. Adapted from Analytical Biochemistry, vol 440, Gilda, J. E., & Gomes, A. V., “Stain-Free total protein staining is a superior loading control to β-actin for Western blots,” pages 186188, copyright 2013, with permission from Elsevier.

In this figure, panels AB show the housekeeping protein approach. By eye, it’s difficult to see an increasing concentration of β-actin in panel A. Indeed, the quantification in panel B is not especially linear.

In panel C, two different methods of quantifying total protein are compared. While a higher slope indicates higher sensitivity, linearity is the most important feature of a loading control, and both total protein measurements are linear. Note the different y-axes between panels B and C, indicating that β-actin is less sensitive as well as less linear than the total protein quantification. This illustrates why total protein measurement is often more accurate as a loading control.

Thanks for joining us in this exploration of western blot normalization. May your blots be clear and your detection range be linear!

Subscribe to the Addgene blog


Resources and References

References

Aldridge, G. M., Podrebarac, D. M., Greenough, W. T., & Weiler, I. J. (2008). The use of total protein stains as loading controls: An alternative to high-abundance single-protein controls in semi-quantitative immunoblotting. Journal of Neuroscience Methods, 172(2), 250–254. https://doi.org/10.1016/j.jneumeth.2008.05.003

Ghosh, R., Gilda, J. E., & Gomes, A. V. (2014). The necessity of and strategies for improving confidence in the accuracy of western blots. Expert Review of Proteomics, 11(5), 549–560. https://doi.org/10.1586/14789450.2014.939635

Gilda, J. E., & Gomes, A. V. (2013). Stain-Free total protein staining is a superior loading control to β-actin for Western blots. Analytical Biochemistry, 440(2), 186–188. https://doi.org/10.1016/j.ab.2013.05.027

Additional Resources on the Addgene Blog

Resources on Addgene.org

 

Topics: Lab Tips, Molecular Biology Protocols and Tips, Antibodies, antibodies 101

Leave a Comment

Sharing science just got easier... Subscribe to our blog