Skip to main content
Lab Grimoire
TW EN
Coffee
Lab Methods

Western Blot Quantification: Densitometry and Normalization

How to quantify a western blot: band densitometry, background subtraction, linear range, loading control vs total protein normalization, and reporting.

Author
CY
Published
Biomedical Principles and Techniques · Part 1 of 4

Western blot quantification gives a relative amount: subtract the background from a band's integrated intensity, divide by the loading control in the same lane, and express the result as a fold of the control group. The number holds up under three conditions: the target band and the loading control both sit in the linear range, the loading control does not change with the treatment, and the conclusion rests on several independent biological replicates.

For the principles of SDS-PAGE, transfer, and antibody detection, see how SDS-PAGE and Western blot work. For bench steps, see the Western blot protocol. This article starts once you have an image and covers densitometry, linear range, normalization, comparison across membranes, and reporting.

Band densitometry step by step

Densitometry turns the brightness of each band in an image into a number. In their 2020 methods article, Pillai-Kastoori and colleagues point out that every stage of a Western blot, from protein extraction to image analysis, adds error, and that quantitative data need validation at each stage. Image analysis is the last stage. It cannot repair errors made earlier, and inconsistent analysis adds another layer of error.

Use the original image

Quantify from the original file exported by the imaging system, such as a 16-bit TIFF. An 8-bit image has only 256 gray levels; a 16-bit image has 65,536 and can resolve far smaller differences in intensity. Screenshots, JPEG compression, and auto-contrast all change pixel gray values. Adjust brightness and contrast only on copies made for figures, never on the file used for quantification.

Lane and band boundaries

Frame each lane first, then frame the target band within each lane. Three rules apply to the boundaries:

  • Give every lane a box of the same width, wide enough to cover the full band.
  • Set the top and bottom of the band box to enclose the whole band without taking in a neighboring band.
  • Use the same box size for the same protein across one membrane.

A band's integrated intensity is the sum of the gray values of all pixels in the box. A larger box takes in more background, so numbers from boxes of different sizes cannot be compared.

Background subtraction

Areas of the membrane without bands still have gray values. This is the background. There are two common ways to subtract it. The first takes a blank region of the same size next to the band and uses its intensity as the background. The second plots the intensity profile of the lane, draws a baseline under the band's peak, and counts only the area above the baseline.

Taylor and Posch (2014) describe background subtraction as a common reason for variable or incorrect Western blot data. With box-based measurement, the background comes from a location outside the band, and the fainter the band, the larger the share of background inside its box. They recommend a lane profile combined with a rolling disc algorithm that subtracts the background within each lane.

The Western blot densitometry tool uses a lane baseline by default: it draws a straight line between the top and bottom edges of the band and integrates only the area above that line. For images with a very flat background, you can switch to the border median. Whichever method you choose, use it for every lane on the membrane. For samples spotted directly onto a membrane without electrophoresis, use the dot blot analyzer, which takes the median of a ring around each dot as the background.

Worked example

The numbers below are illustrative, not measured data. The three lanes are control, treatment A, and treatment B. Each lane has two bands, the target protein and the loading control, measured in arbitrary units.

Lane Target: raw − background = net Loading control: raw − background = net Target ÷ loading control Fold of control
Control 1,520 − 320 = 1,200 4,350 − 350 = 4,000 0.300 1.00
Treatment A 2,310 − 330 = 1,980 4,680 − 360 = 4,320 0.458 1.53
Treatment B 980 − 310 = 670 4,010 − 340 = 3,670 0.183 0.61

The calculation has three steps. First, subtract the background from the raw integrated intensity to get the net intensity. Second, divide the net target intensity by the net loading control intensity from the same lane. Third, divide each lane's ratio by the control lane's ratio of 0.300, which sets the control to 1.00.

The numbers show what the loading control does. The net loading control intensity in treatment A is 1.08 times that of the control, so this lane received about 8% more protein. Without dividing by the loading control, the net target intensities alone would give 1.65-fold for treatment A and 0.56-fold for treatment B. After dividing, the values are 1.53-fold and 0.61-fold.

Linear range and saturation

The linear range is the interval in which signal intensity is proportional to the amount of protein. Only within this interval does a doubling of band intensity mean a doubling of protein.

Signal leaves the linear range in two ways. The first is in the image: once a pixel reaches the maximum gray value it cannot go higher. This is saturation, and the integrated intensity of a saturated band is underestimated. The Western blot densitometry tool flags bands that contain saturated pixels with a warning mark in the results table; when you see it, use an image with a shorter exposure. The second is on the membrane and in antibody binding: at high protein loads, the signal stops rising in proportion to the amount of protein even when no pixel is saturated. Taylor and Posch note that the common loads of 10–100 µg per lane usually have no experimental basis, and that abundant housekeeping proteins are often already outside the linear range at these loads.

A dilution series tests the linear range. Taylor and Posch recommend pooling all samples, loading a twofold serial dilution of the pool, and plotting band intensity against protein load. The linear range is the interval in which the intensity halves each time the load halves, and the load for the actual experiment should fall in the middle of it. Check each antibody separately, because the target protein and the loading control usually have different linear ranges.

The lower limit of the linear range needs the same attention. When a band is faint and close to the background, small differences in background subtraction cause large relative errors. Load more protein or expose longer, then acquire a new image.

Exposure time follows the same principle. Capture a series of exposures of the same membrane from short to long, and quantify from the one in which neither the target band nor the loading control has saturated pixels. When the target and the loading control differ greatly in signal, each can come from a different exposure, but all lanes of the same protein must come from the same image.

Normalization: housekeeping protein or total protein

Normalization corrects for differences in protein load and transfer efficiency between lanes. There are two choices for the denominator: a single housekeeping protein, or the total protein in the lane.

When housekeeping proteins fail

β-actin, GAPDH, and α-tubulin are the most widely used housekeeping proteins. Their use as loading controls rests on two assumptions: band intensity changes in proportion to the protein load, and their abundance is unaffected by the treatment. Published studies document cases in which each assumption fails.

  • The signal does not follow the load. Dittmer and Dittmer (2006) found that at the higher protein loads needed to detect low-abundance proteins, β-actin antibodies could not distinguish differences in actin levels; diluting the antibody or changing the incubation time helped little. Aldridge and colleagues (2008) reported a similar result: at the protein concentrations normally used to quantify low-abundance proteins, GAPDH and β-actin did not accurately reflect differences in protein concentration.
  • Abundance changes with conditions. Greer and colleagues (2010) reported that the density of cultured cells significantly affected the levels of α-tubulin and GAPDH, while HSP90 and β-actin did not change in the same experiments. This result applies to the cells and conditions of that study; a different experimental system needs its own check.

The first case is easy to miss. When the loading control is outside the linear range, its bands look equally dark in every lane, which resembles even loading. In fact it can no longer report differences in load.

Total protein normalization

Total protein normalization uses the total protein stain signal of the whole lane as the denominator. Stains include Ponceau S, Coomassie Blue, Amido Black, SYPRO Ruby, and stain-free gels. Aldridge and colleagues compared SYPRO Ruby and Amido Black and found both to be acceptable alternatives to single-protein controls.

The total protein signal is the sum of many bands in the lane, so it does not depend on the stable expression of one protein. Taylor and Posch report that stain-free technology is linear between about 10 and 80 µg of protein load from a typical cell or tissue lysate. Both authors work for the company that makes this technology, so state the source when you cite this range. A total protein stain also has a linear range and needs a dilution series to confirm it.

Choosing and validating

If you use a housekeeping protein, show two things with your own samples and protein loads: the loading control signal is within the linear range, and the amount of loading control does not differ between groups. Aldridge and colleagues conclude that the burden of proof lies with the researcher to show that a loading control reflects differences in protein amount. Confirm the first point with a dilution series. For the second, compare the net loading control intensity with the total protein stain signal in each group; the ratio between the two should be the same across groups. If you cannot show both, use total protein normalization.

Comparing membranes and biological replicates

Comparisons between membranes

Transfer efficiency, antibody incubation, and exposure differ from one membrane to the next, so raw intensities from different membranes cannot be compared directly. Three practices help:

  • Put the samples you intend to compare on the same membrane whenever possible.
  • Take the target protein and its loading control from the same lane of the same membrane.
  • When the samples do not fit on one membrane, load the same reference sample on every membrane and convert each membrane's values relative to that sample. The workflow described by Taylor and Posch places the same control sample in the first lane of every membrane.

Number of biological replicates

Samples from independent cultures, independent animals, or independent treatments are biological replicates. Loading the same lysate again, or measuring the same membrane again, gives technical replicates, which reflect only handling and measurement error.

A statistical test needs an estimate of within-group variation, which cannot be calculated from one sample per group. Three independent biological replicates is a common minimum in the literature, but three samples can detect only large differences. When you expect a smaller difference, estimate the sample size from the expected variation before the experiment.

Comparing groups

To compare two groups, enter the target ÷ loading control ratio from each independent experiment into the t-test calculator. When every membrane carries both a control and a treated sample, the design is paired, so choose the paired t-test. Setting the control on every membrane to 1.00 and then running an independent-samples test is not appropriate, because it artificially removes all variation from the control group. The bar chart in the Western blot densitometry tool compares each group with the control by Welch's t-test with Holm correction. Each group needs at least two lanes, and whether those lanes are biological or technical replicates depends on your design.

What journals ask authors to report

Take the Journal of Biological Chemistry (JBC) as an example. In a 2015 editorial, Fosang and Colbran set out the journal's requirements for immunoblot data. Other journals have different rules, so read the author guidelines of your target journal before you submit. The JBC requirements can be summarized as follows:

  • Quantification method. Describe how signal intensity was measured and how protein loading was normalized between lanes. Quantification of Western blots is not always required. The editorial notes that quantification depends on a linear relationship between signal intensity and the amount of sample loaded, and that chemiluminescence recorded on X-ray film has a very restricted linear range.
  • Denominator for normalization. JBC prefers normalization to total protein staining of the membrane and cautions against housekeeping proteins unless the authors show that the housekeeping protein's expression is unaffected by the experimental conditions.
  • Replicates. State separately the number of independent biological replicates, the number of technical replicates, and the number of repeated experiments.
  • Image integrity. Do not splice images without disclosure. Mark each splice position and explain it in the figure legend. Lanes from different blots must not be spliced together. Avoid overcropping, and keep the position of at least one molecular weight marker above and one below the band of interest.
  • Antibodies. List the species of origin, supplier, catalogue number, and lot number of each antibody, and provide data that support its specificity.
  • Raw data. Keep the original images; the journal may ask for them during review.

Record every setting in the quantification workflow, including the image file format, box size, background subtraction method, and normalization denominator, and report them in the methods section.

References

  • Pillai-Kastoori L, Schutz-Geschwender AR, Harford JA. A systematic approach to quantitative Western blot analysis. Anal Biochem. 2020;593:113608. doi:10.1016/j.ab.2020.113608
  • Taylor SC, Posch A. The design of a quantitative western blot experiment. Biomed Res Int. 2014;2014:361590. doi:10.1155/2014/361590
  • Aldridge GM, Podrebarac DM, Greenough WT, Weiler IJ. The use of total protein stains as loading controls: an alternative to high-abundance single-protein controls in semi-quantitative immunoblotting. J Neurosci Methods. 2008;172(2):250–254. doi:10.1016/j.jneumeth.2008.05.003
  • Dittmer A, Dittmer J. β-Actin is not a reliable loading control in Western blot analysis. Electrophoresis. 2006;27(14):2844–2845. doi:10.1002/elps.200500785
  • Greer S, Honeywell R, Geletu M, Arulanandam R, Raptis L. Housekeeping genes; expression levels may change with density of cultured cells. J Immunol Methods. 2010;355(1–2):76–79. doi:10.1016/j.jim.2010.02.006
  • Fosang AJ, Colbran RJ. Transparency is the key to quality. J Biol Chem. 2015;290(50):29692–29694. doi:10.1074/jbc.E115.000002

References

  1. A systematic approach to quantitative Western blot analysis
  2. The Design of a Quantitative Western Blot Experiment
  3. The use of total protein stains as loading controls: An alternative to high-abundance single-protein controls in semi-quantitative immunoblotting
  4. β-Actin is not a reliable loading control in Western blot analysis
  5. Housekeeping genes; expression levels may change with density of cultured cells
  6. Transparency Is the Key to Quality

Frequently Asked Questions

How is a western blot quantification value calculated?

Subtract the background from the band's integrated intensity to get the net intensity, divide by the net loading control intensity in the same lane, then divide by the control group's ratio to get the fold of control. The result is a relative amount, not an absolute concentration.

What is wrong with using a housekeeping protein as the loading control?

It can fail in two ways. First, β-actin and GAPDH are abundant, and at the protein loads needed to detect low-abundance proteins their signal is often outside the linear range and no longer reflects differences in load. Second, housekeeping protein levels can change with experimental conditions; for example, the density of cultured cells affects α-tubulin and GAPDH. Validate the control with your own samples before using it.

What is total protein normalization?

It uses the total protein stain signal of the whole lane as the denominator instead of a single housekeeping protein. Stains include Ponceau S, Coomassie Blue, Amido Black, SYPRO Ruby, and stain-free gels. The signal is the sum of many bands, so it does not depend on the stable expression of one protein, but its linear range still needs to be confirmed.

Can western blot results from different membranes be compared directly?

Raw intensities cannot be compared directly, because transfer efficiency, antibody incubation, and exposure differ between membranes. Put the samples you intend to compare on the same membrane whenever possible. When they do not fit on one membrane, load the same reference sample on every membrane and convert each membrane's values relative to that sample.

Found this useful? RSS · X · Buy me a coffee