A systematic approach to quantitative Western blot analysis

Contents lists available at ScienceDirect

Analytical Biochemistry

journal homepage: www.elsevier.com/locate/yabio

A systematic approach to quantitative Western blot analysis

Lakshmi Pillai-Kastoori∗, Amy R. Schutz-Geschwender, Jeff A. HarfordLI-COR Biosciences, 4647 Superior Street, Lincoln, NE, 68504, USA

A B S T R A C T

Attaining true quantitative data from WB requires that all the players involved in the procedure are quality controlled including the user. Appropriate proteinextraction method, electrophoresis, and transfer of proteins, immunodetection of blotted protein by antibodies, and the ultimate step of imaging and analyzing thedata is nothing short of a symphony. Like with any other technology in life-sciences research, Western blotting can produce erroneous and irreproducible data. Weprovide a systematic approach to generate quantitative data from Western blot experiments that incorporates critical validation steps to identify and minimizesources of error and variability throughout the Western blot process.

1. Introduction

Western blotting is a simple yet powerful procedure to investigatethe presence, relative abundance, relative mass, presence of post-translational modifications (PTM) as well as to study protein-proteininteractions. These applications of Western blot provide valuable in-formation in both academic research, diagnostic and therapeutictesting. This workhorse method, first described by Towbin et al. [1] andBurnette [2], relies on the specific interaction of antibodies with targetantigens present in the sample mixture. After sample proteins are se-parated on a protein gel and transferred to the membrane, primary andsecondary antibodies are used to bind and visualize the target protein.

The Western blot was originally intended to provide a yes/no an-swer about the presence of the target protein in a protein sample. Thisqualitative method confirms the presence of target bands by simplevisual assessment [3–8]. With the surge in the field of systems biologyand the need to understand complex biological systems, Western Blot isno longer limited to generating qualitative data.

The reproducibility of Western blot analysis and other im-munoassays is an ongoing source of concern in the scientific community[6,9–11]. The immunoblotting process involves a complex series ofinterdependent steps that are influenced by subjective choices and userexpertise [6,11–14]. Variations in experimental design, methodology,and technique can be substantial sources of error. This variability isparticularly troubling for quantitative analysis of Western blot data,where an error could lead to misinterpretation of data [6,14,15]. See-mingly minor or insignificant differences in the reagents and para-meters used for each experiment may have a surprisingly strong influ-ence on the results [6,9,13,14,16–18]. Careful experimental design andwell-characterized methods are essential to avoid common pitfalls anddetermine which assay parameters are most critical for reproducible

results [6,11–13].However, reproducibility does not guarantee accuracy; an assay

may produce precise data that are inaccurate [13,18,19]. Accuracy andprecision are important but distinct aspects of reproducibility. Accuracyindicates how closely the measured values represent the true value ofthe target or analyte in a sample. Do these values reflect a genuinechange or difference in the experimental sample? Precision describesthe repeatability, variation, and error of assay measurements. If the testis repeated many times, how similar are the results? quantitative ana-lysis of Western blot data should strive for both precision and accuracy,to help ensure that the reported observations convey meaningful in-formation about the experimental samples [14,19–22].

This review aims at providing a step-by-step breakdown of the ap-proach to perform and gather quantifiable data from Western blots(Fig. 1).

2. Quantitative Western Blot analysis: core concepts

At the heart of protein detection via Western Blotting is the use ofquality reagents and correct methodology. While there are several re-sources available to overcome the technical limitations of WesternBlotting, there is limited information available on how data can bequantified and the correlation between the combined linear range ofdetection and normalization strategy.

2.1. Break it right: “Fit for purpose”

The quality of the data gathered from a Western blot is as good asthe quality of the sample used. Protein extracts need to be prepared,purified and quantified using reagents that best suit the sample type.How would you choose from the numerous extractions, fractionation,

https://doi.org/10.1016/j.ab.2020.113608Received 1 August 2019; Received in revised form 16 December 2019; Accepted 27 January 2020

∗ Corresponding author.E-mail addresses: [email protected], [email protected] (L. Pillai-Kastoori).

Analytical Biochemistry 593 (2020) 113608

Available online 31 January 20200003-2697/ © 2020 LI-COR BIOSCIENCES, LINCOLN, NEBRASKA 68504. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/BY-NC-ND/4.0/).

T

http://www.sciencedirect.com/science/journal/00032697

https://www.elsevier.com/locate/yabio

https://doi.org/10.1016/j.ab.2020.113608

https://doi.org/10.1016/j.ab.2020.113608

mailto:[email protected]

mailto:[email protected]

https://doi.org/10.1016/j.ab.2020.113608

http://crossmark.crossref.org/dialog/?doi=10.1016/j.ab.2020.113608&domain=pdf

and purification methods? Several comprehensive reports have beenpublished by researchers including Bass et al. [23], Pandey et al. [24],Wingfield et al. [25], Peach et al. [26], Gavini et al. [27], and Miskiewizet al. [28], describe in great details various considerations for preparingquality protein extracts. Work published by Murphy and the group alsodescribes the implication of sample fractionation on downstream pro-tein detection [29]. Users' experimental workflow influences thestrategy for extraction and solubilization of target protein. It is critical,however, to standardize the sample preparation method in alignmentwith the experimental design and using consistent protocols throughoutthe duration of the study. In summary, the user needs to consider thetype of target protein (nuclear, transmembrane, mitochondrial, etc.) aswell as the type of tissue/cell that houses the target in order to choosethe extraction, purification, and quantification strategy.

2.2. Sample: loaded or overloaded?

The next important question is; how much sample needs to beloaded for separation using two-dimensional electrophoresis. Sodiumdodecyl sulfate-polyacrylamide gel (SDS-PAGE) matrix, electroblotting(transfer), and protein binding capabilities of membranes influencetogether with the type of target protein influence the overall perfor-mance of Western Blotting technique. Overloading of samples is awidespread problem that often goes unrecognized and may compromisethe accuracy of quantitative analysis [3,4,6,7,30,31]. Murphy andothers have successfully shown that in fact, loading low amounts of

protein sample can improve outcomes in Western Blotting [15,29,32].For accurate quantitative analysis, sample loading must be evaluatedand calibrated experimentally by the user. How do you determine thecorrect loading concentration for the sample of interest?

For accurate quantitative data analysis, the relationship betweensample loading and band intensity must be evaluated and calibrated todetermine the linear range of detection for the assay [3,6,7,29]. Thelinear range of detection is the range of sample loading where bandintensity increases in proportion to sample loading or target abundance.Within this range, a change in sample input will produce a linear andproportional response in signal output which is determined coefficientof determination R2. Closer the R2 value is to more the linear regressionreflects the data. For example, a two-fold or five-fold increase in sampleloading should theoretically induce an equivalent increase in relativefluorescence units which would be recorded as band intensity (Fig. 2).Analyzing data outside of the linear range may result in inaccurate andnon-reproducible data. All quantification must be performed within therange where an increase in sample input (x, 2x, and 5x) produces alinear and proportional response in signal output (y, 2y, and 5y). Abovethe linear range of detection (shoulder), strong bands exceed the ca-pacity of the assay and become saturated; the expected, proportionalincrease in band intensity does not occur and strong signals will beunderestimated. Below the linear range (tail end), faint bands are dif-ficult to reliably distinguish from membrane background and do notreflect actual differences in band intensity.

We strongly emphasize the dual importance of linearity and pro-portionality in the quantitative analysis of Western blot data. Linearregression can be used to fit a data set to a straight line, but this doesnot imply or guarantee proportionality. An R2 value equal to 1 indicatesthat the regression line represents the data perfectly, but this line mayor may not represent a proportional signal response. A linear andproportional relationship is represented by the equation y = mx, whichdescribes a straight line passing through the origin. However, it may bepossible to identify regions or subsets of the data that do display aproportional response.

2.3. Combined linear range (CLR) of detection

The linear range of detection is central to quantitative analysis of

Fig. 1. Quantitative approach to Western Blotting. Understanding of thephysiology and cellular location of the target protein will allow for optimalprotein extraction outcomes from various sample source. Quantification ofprotein extracts prior to Western blotting is crucial first step in generation ofaccurate data. Validation of antibodies is required both by the vendor as well asthe end-user in the context of the experimental system under study. Internalloading controls (ILC) needs to identify and validated within the context of theexperimental study. Both target and internal loading control needs to detect andanalyzed within the combined linear range (CLR) of detection. Normalization ofthe target protein to appropriate ILC will provide an accurate representation ofhow target protein abundance is affected in the experimental study. Finally,biological and technical replicates not only provide statistical power but alsorule out user bias and accounts for biological variability.

Fig. 2. Linear relationship between sample concentration and band in-tensity. The signal derived from the protein bands on a Western blot varieswith the amount of sample extract loaded onto the protein gel. The illustratedgraph depicts a linear and proportional relationship between amount of sampleloaded (x, 2x, 5x) and the relative fluorescence units (R.F.U) captured from thetarget bands (y, 2y, 5y). Tail and shoulder end of the data curve capture noiseand saturated signal, respectively. μg, micrograms; R.F.U, Relative fluorescenceunits.

L. Pillai-Kastoori, et al. Analytical Biochemistry 593 (2020) 113608

2

Western blot data; the accuracy and validity of analysis are based onthis foundation. Because it is influenced by numerous experimentalfactors, the linear range should be determined empirically for eachWestern Blot assay. A combined linear range denotes the linear rangefor both the target protein as well as a loading protein [also calledInternal Loading Control (ILC)]. Internal loading control as the namesuggests is present internal to the test sample itself and could be a singleHousekeeping protein (HKP) or total cellular protein. Serial dilutions ofbiological samples can be used to identify the appropriate linear rangeof sample loading for both the target protein(s) and internal loadingcontrol. It is important to note that this linear range will likely be dif-ferent for each target protein and loading control combinations andmay be affected by the detection chemistry and/or imaging platform.The relationship between sample input and band intensity must be bothlinear and proportional for both target protein and internal loadingcontrol to enable quantitative analysis. Choosing an appropriate ILCmay seem simple, but it is an important and sometimes obscure aspectof experimental design for accurate quantitative analysis of Westernblot data [6,10].

2.3.1. Internal loading controlsThe ILC must meet the two requirements when evaluated in the

specific conditions, treatments, cell or tissue types, and experimentalcontext where it will be used. Because experimental manipulations mayinfluence the expression of the ILC, biological stability should be vali-dated in all samples that represent the intended conditions and treat-ments. Biological variability in expression of the ILC is a source of errorthat may undermine the accuracy of the analysis; particularly for ana-lysis of small differences in target abundance [3,11,33–35]. Thevariability introduced by technical limitations (gel matrix, transferconditions, buffers, etc.) of the Western blot method should also beconsidered when choosing and validating an ILC. Either type of varia-bility may make it more difficult to reliably detect small differences andeffects, leading to false-negative results [4,15,36,37].

Normalization of a target protein is most accurate when the targetprotein and ILC are detected in the same lane on a single blot[6,8,38,39]. To achieve this goal, detection must be performed in thecombined linear range – the range of sample loading that produces alinear, proportional response in band intensity for both the target andthe ILC [3,6,7,29,33]. If they cannot be accurately detected in the samerange of sample loading, the results will not be meaningful; a differentILC should be selected for the experiment. This may occur for a varietyof reasons, such as the use of an abundant HKP as an ILC for a low-abundance target, a pan/PTM analysis experiment where an abundanceof unmodified and modified forms varies dramatically, if primary an-tibody specificity and affinity are poor, or if transfer methods requirefurther optimization.

2.4. Internal loading control: housekeeping protein

A single endogenous reference protein, such as a housekeepingprotein, is often used as an ILC. Members of Actin, tubulin families andglyceraldehyde‐3‐phosphate dehydrogenase (GAPDH) are common ex-amples. This method is referred to in the literature as a single-proteinloading control [3]. Although this method is widely used and can bequite accurate, it is uniquely vulnerable to biological variability [40].An HKP or other single-protein loading control alters and reformulatesthe experimental hypothesis, as described by Aldridge et al. [3]. Ratherthan evaluating the abundance of the target protein relative to totalsample protein or cell number, the experiment now examines targetabundance relative to that particular single-protein loading control [3].If the expression of that single protein is stable and unaffected by therelevant experimental conditions and samples, this ILC meets the firstrequirement and the hypothesis is valid. But if the expression of thatsingle-protein loading control is variable, the reformulated hypothesiswill fail.

Adoption of HKPs as ILCs for immunoblotting was inspired byhousekeeping genes, the stably expressed transcripts used as en-dogenous controls for analysis of gene expression analysis. Stable ex-pression of actin, tubulin, and other housekeeping genes in many celltypes and tissues initially indicated that their protein products shouldbe similarly appropriate for Western blot normalization. However, theexpression of some HKPs is now known to be altered in response tocertain experimental conditions. Recent studies have demonstratedregulated expression of actin, tubulin, and other HKPs in response toconditions such as cell confluence, disease state, developmental stage,hormonal state, drug treatment, and cell or tissue type[33,35,37,41–49]. Manceau et al. [43], demonstrate the variability inβ-tubulin and β-actin across different tissues extracts from mouseneonates. Because the variability of HKP expression in experimentalsamples will introduce error in quantitative analysis, validation is animportant first step. Before an HKP or other single-protein loadingcontrol is used for quantitative analysis of Western blot data, its stableexpression should be demonstrated and verified for the relevant ex-perimental samples and manipulations [3,29,33–35,37,47].

2.5. Internal loading control: total cellular protein

In recent years, total cellular protein staining of the blotted mem-brane has emerged as the preferred method for normalization of sampleloading in quantitative analysis of Western blot data[35,37,41,44,47,50–53]. Prior to the immunodetection of the blot, afluorescent total protein stain is used to visualize and assess actualsample loading across the blot. The membrane is then processed forantibody incubations and detection. Total protein loading controls aremuch more resistant to biological variability than an HKP or othersingle-protein loading control [3,15,31,35,44,47,50,51] and may pro-vide a wider linear range of detection than an HKP – and because of itsresistance to biological variability, it requires less validation prior touse [3,4,7,15,35,41,44,50,51]. Because variation is often introducedduring electrophoretic transfer from the gel to a membrane, total pro-tein staining of the blot is preferable to staining of the gel prior totransfer or staining of a duplicate gel loaded with the same samples[3,35,47,51]. Any method used for the detection of total proteinloading should not damage the bound sample proteins or interfere withsubsequent antibody-based detection of the blot [6,31,35,54,55]. Atotal protein stain should meet a number of criteria, including an ap-propriate linear range of detection that is compatible with the target ofinterest; stable signals and low sample-to-sample variability; lack ofinterference with immunodetection and other downstream analysis;suitability for membrane staining, to correct for variability duringtransfer; and compatibility with multiplex detection and quantification[3,6,35,41,47,51–53,55,56]. Several fluorescent membrane stains haverecently been described in the literature, including Ponceau Stain [55],Coomassie Brilliant Blue, SYPRO Ruby stain, Blot FastStain [57], andRevert 700 Total Protein Stain [35,58–62].

2.6. Detection chemistry

Two main types of detection chemistry are used to visualize signalson Western blots. Enhanced chemiluminescence (ECL) is a popularenzymatic method that uses horseradish peroxidase (HRP) as an in-direct reporter of secondary antibody binding. This HRP reporter pro-duces photons of light as it consumes a luminol-based substrate.Fluorescence detection is a non-enzymatic method that uses fluor-ophore-labeled secondary antibodies as a direct reporter of antibodybinding. Although chemiluminescence is widely used, its inherent en-zyme/substrate kinetics are a source of variability and often fail toproduce the linear, proportional signal response required for quantita-tive analysis of Western blot data [5,6,8,11,63]. The rate of the che-miluminescent reaction is dependent on the local concentration of en-zyme and substrate. Substrate availability is a dynamic property, and


3

the reaction rate will vary continuously over time and across the surfaceof the blot. A low-intensity band will consume the available substratemuch more slowly than a strong band with a high local concentration ofHRP reporter that may rapidly deplete the available substrate. If theHRP concentration is too high, a signal may be lost in areas where thesubstrate is rapidly consumed and exhausted; this is the cause of ghostbands (reverse banding) and “burned-in” bands with brown or yellowprecipitate [6,64]. Different chemiluminescent substrate formulationcan be specifically used to detect femtograms (ultrasensitive), pico-grams or milligrams of target protein within a lysate [65,66]. As suchsimultaneous detection of high and low concentration of proteins in thesame blot can be challenging with one formulation ECL substrate[67–69] thereby, resulting in a non-linear and narrow dynamic range ofdetection. However, titration of protein amount, the primary antibodycan help in mitigating all the above-mentioned issues, provided thescientist is cognizant [52,70].

We and others developed fluorescent Western blot methods as amore convenient and quantitative alternative to indirect chemilumi-nescence [63,71]. Fluorophores are retained at the site of antibodybinding and emit photons upon exposure to the appropriate wavelengthof excitation light, typically in the near-infrared (NIR) spectrum[5,6,8,38,71–73]. Signals are very stable and much more reproduciblebecause they are unaffected by timing, substrate availability, and othervariables that limit the usefulness of chemiluminescent methods forquantitative analysis [5,6,8,11,74]. Fluorescence imaging also enablesmultiplex detection of a target protein and ILC in the same lane on thesame blot, for more accurate correction of sample-to-sample and lane-to-lane variation [6,8,38,63]. Stripping of Western blots for re-probingcan cause substantial loss of sample proteins from the membrane and isan avoidable source of error [6,8,30,31]. Fluorescence is generally re-cognized as the most accurate and reliable method for quantitativeanalysis of Western blot data [4–8,41,73,74]. Source of species in whichantibodies were raised, the wavelength of secondary antibody taggedfluorophore, and imaging system needs are critical to performingfluorescence-based Western blot detection of proteins.

Saturation artifacts are not readily apparent in a Western blot imageand may go undetected if serial dilutions of a sample are not examinedto define the linear range of detection. Two types of saturation arefrequently observed in quantitative analysis of Western blot data.Membrane saturation is caused by the overloading of the sampleprotein. During the transfer of an overloaded gel to a blotting mem-brane, abundant sample proteins bind in layers on the surface of themembrane [11,15,31,76]. If highly abundant proteins exceed the localbinding capacity of the membrane, they will be washed away. Thelayering of abundant proteins also limits antibody access and may onlyallow antibodies to bind the top layer of protein. This layering effectcauses underestimation of strong bands and may also interfere with thedetection of low-abundance proteins on the blot. Membrane saturationis common may be difficult to identify without the evaluation of a di-lution series of the sample. This type of saturation can affect anyWestern blot assay, regardless of the detection chemistry or imagingmethod [11,15,77].

Signal saturation occurs when the intensity of a strong band ex-ceeds the capacity of the detection chemistry or imaging system. At thispoint, strong signals begin to plateau and increasing amounts of targetno longer produce the expected increase in signal (Fig. 3). Chemilu-minescent detection is highly susceptible to saturation; the generationof a signal is restricted by substrate availability, and the linear range isquite limited, even when detected by digital imaging [5–7,11,15,74].Fluorescent Western blot methods are relatively resistant to saturationand provide a much broader linear range, particularly when imagedwith a digital system that provides an appropriately wide dynamicrange [5–8,73,74]. However, in heavily loaded samples, self-quenchingof tightly packed fluorophores may be possible [6].

2.7. Normalized Western blot data

Normalization mathematically corrects for small, unavoidable var-iations in sample loading and protein transfer by comparing the targetprotein to the ILC in each lane. The target protein signal in each lane isdivided (normalized) by the signal value for the ILC in that lane. Whentarget signals are normalized to sample loading, relative levels of thetarget protein can be compared across the blot to determine if changesin band intensity represent biological differences between samples.Normalization is based on the fundamental assumption that both thetarget and ILC signals are dependent on sample concentration. For thisassumption to hold true, as mentioned above the ILC must meet tworequirements: it must be expressed at a generally constant level acrossall relevant experimental samples and conditions; and the resultingsignal intensity must be proportional to abundance, without saturationeffects or technical limitations. A loading control that does not meetboth requirements is not an accurate indicator of sample loading andshould not be used for normalization and quantitative analysis ofWestern blot data [3,31,47,51].

Without normalization, an apparent difference in target abundanceon a Western blot cannot be accurately interpreted. This concept is il-lustrated in Fig. 4. Detection of the target protein is seen in Fig. 4A(green). Plotting the raw intensity values for the target may suggest thattarget abundance is variable among the different samples (Fig. 4C).However, the observed difference between target bands may be theresult of inconsistent loading of total protein extract (Fig. 4B). Sample#4 appears to have reduced levels of target protein compared to rest ofthe samples, however, after normalization to correct variations insample loading; it is evident that sample extracts were loaded unequallyin each lane and therefore, any changes in target protein observed maybe due to improper loading of protein extracts and not due to experi-mental intervention (Fig. 4D).

Normalization is appropriate for the correction of small and un-avoidable differences in cell number, sample concentration, loadingerror, and position or edge effects from electrophoresis and transfer.However, it should not be relied on to correct for preventable error andvariability in the quantitative analysis of the Western blot process [3,6].Normalization should be used in addition to careful experimental de-sign and sample preparation, not in place of it. Minimizing or elim-inating sources of variation throughout the experiment is critical forreproducible quantitative analysis of Western blot data (see Refs.[6,7,29,78] for excellent information about protein extraction, samplehandling, and other error-prone aspects of the Western blot method).

Total protein normalization is an aggregate method that combinesthe band intensities from many different sample proteins in each lane –essentially using multiple ILC proteins to provide a more accuratereadout of sample loading. This multi-protein approach is also callednormalization by sum [36]. This normalization method also offers amore direct readout of the cellular material loaded than antibody-baseddetection of a single endogenous protein [3,4,31,35,36,41,44]. Totalprotein normalization is now increasingly viewed as the “gold stan-dard” for performing quantitative analysis of Western blot data [52,53].The Journal of Biological Chemistry recommends normalization to totalprotein loading, clearly stating their preference “that signal intensitiesare normalized to total protein by staining membranes with CoomassieBlue, Ponceau S, or other protein stains” [52,53].

3. Background subtraction

Western blots not only contain the signal generated by the im-munodetection of the target protein but also signal generated as a resultof nonspecific bands, smears, and the inherent signal produced by themembrane itself. When a target band is quantified the backgroundsignal around the band is also included in the data analysis, and thegoal is to reduce/remove the confounding error introduced by suchbackground signal [23,79,80]. Currently, several Western blot analysis


4

software packages offer multiple options to subtract background signaland quantify target bands. Each Western blot has a unique profile withrespect to artifacts, uniformity of background, positioning of lanes andbands and requires a background subtraction algorithm that can adaptto the said variations.

There are several methods to subtract the background from theWestern Blot image [23,79,81–84] either using the Shape-Basedmethods for local and/or global background subtraction; Lane-Basedmethods such as Rolling Ball subtraction. A new patent-pendingmethod called Adaptive Background Subtraction (ABS) removes user-bias by taking into consideration both shape as well as lane profiles togenerate reproducible and accurate quantitative Western Blot data. ABSbackground subtraction algorithm present in Empiria Studio Software

works well with real-life Western blots that have oddly-shaped mis-shapen bands and smears, and the data generated has greater accuracythan local and global subtraction methods (For more details read [85];Fig. 5).

4. A systematic approach to accurate quantitative analysis ofWestern blot data

Here, we describe a five-step systematic strategy to increase theprecision and reliability of the quantitative data generated from theWestern blotting technique. This process incorporates critical valida-tion steps that address common, but sometimes unrecognized, sourcesof error and variation. Because we and others strongly prefer and

Fig. 3. Saturation of strong bands obscures actual differences between samples. Increasing amounts of C32 (sc-2205) whole cell lysate (1–60 μgs) were loadedon 4–12% Bis-Tris PAGE gels and transferred onto a PVDF membrane. Fluorescent detection of Actin was detected on three replicate blots performed on differentdays. A) Actin bands were analyzed and quantified with Image Studio™ analysis software (LI-COR Biosciences) and the detected band intensities compared to thepredicted signal intensity, based on the known amount of sample protein loaded in each lane. Results of three replicate experiments are shown (blue line indicatesmean values; dotted line shows predicted band intensity for a linear and proportional response. Above 10 μg of cell lysate per lane, actin signals begin to plateau, andincreased sample loading does not produce the expected increase in band intensity. B) Differences between high-intensity bands were underestimated by quantitativeanalysis. Pink bars show the actual mean band intensity of replicate blots [± Standard Deviation (STDEV.)]; blue bars indicate the predicted increase in bandintensity. A comparison of the 5 μg and 20 μg sample lanes indicates a 3.1-fold increase in signal, lower than the predicted 4-fold increase. Comparison of the 10 μgand 30 μg sample lanes indicates a larger discrepancy in band intensity: a 1.6-fold increase is observed, roughly half of the expected 3-fold change. At or above 30 μgof sample, increases in band intensity are minimal and differences between lanes are not reproducibly detected (saturated). (For interpretation of the references tocolour in this figure legend, the reader is referred to the web version of this article.)

Fig. 4. Normalization enables accurate interpretation of differences in protein expression. This is a simulated Western blot representing unequal loading ofsample extracts to visualize a target protein and total protein extract via antibody and Total Protein Stain respectively. Total protein will be used as an ILC in thesimulated data analysis. A) This illustration depicts Western blot results for a target protein in lanes 1–5 (green bands). Lanes 3 and 4 appear to have higher signal forthe target protein compared to lane 4. B) The same blot from panel A is presented to depict the total protein extract in lanes 1 through 5 (red). Lanes 3 and 4 appear tohave higher total protein extract compared to lane 4. C) Raw intensity values for the target protein bands plotted against the sample number suggest varyingexpression levels of the target protein. D) Upon normalization of the target protein bands in panel A with the total protein signal in panel B it is evident that sampleextracts were loaded unequally in each lane and therefore, any changes in target protein observed may be due to improper loading of protein extracts and not due toexperimental intervention. Such a Western blot would not provide accurate and reproducible qualitative data. R.F.U; Relative fluorescence units. (For interpretationof the references to colour in this figure legend, the reader is referred to the web version of this article.)


5

recommend fluorescent Western blot methods for quantitative analysis[5–8,52,63,74], this strategy was specifically developed for use withfluorescence-based detection.

4.1. Validate the primary antibody

After sample proteins are separated on a protein gel and transferredto the membrane, primary and secondary antibodies are used to bindand visualized the target protein. This process relies on two key prop-erties of the primary antibody: specificity, the antibody's ability to re-cognize and bind to the target antigen; and selectivity, the antibody's

preference to bind the target antigen in the presence of a heterogeneousmixture of sample proteins. These characteristics should be verified andvalidated for each primary antibody. Because antibody performance isgreatly influenced by the assay context and parameters, validationshould be performed by the user in the intended assay and relevantexperimental context [18,86–94]. The reproducibility of results pro-duced with the primary antibody should also be verified [18,76,88,95].

4.1.1. SpecificityThe specificity of antibody within the realm of a Western blot means

that the antibody recognizes the target protein(s), either as a single

Fig. 5. Quantitative analysis with Adaptive Background Subtraction (ABS) is more reproducible than local and global background subtraction. Multipleusers analyzed the designated bands on each blot (arrowheads on images) using local, global/user-defined, and Empiria Studio ABS background subtraction.Variability of background-corrected band intensity, expressed as the mean % CV of the quantitative results for that image, was compared for each backgroundsubtraction method. A) Blot A displayed oddly shaped bands and smeared trails in sample lanes. Graph shows the mean CV of quantitative analysis for eachbackground method (error bars show standard deviation, n = 6). B) On Blot B, faint bands were surrounded by background in sample lanes. Graph shows the meanCV for each method (error bars show standard deviation, n = 6). C) The designated bands on Blot C were smeared and misshapen. Graph shows the mean CV for eachmethod (error bars show standard deviation, n = 5).


6

distinct band or a set of bands of the correct molecular mass. Detectionof a single band at the expected molecular weight is an important firststep but is not sufficient to prove antibody specificity. It is wise to re-member that the presence of “one band” on a blot does not mean “onetarget” [7,53,86,87,96,97]. Antibodies can be validated by testing theperformance of antibodies on either genetic knockout, knockdown,positive, and negative protein samples (see for more details[92,98–104]). The specificity of the antibodies against Pan andPhospho-EGFR was validated in a 2-pronged approach by using apeptide blocking strategy. A single blot (Fig. 6A and B) is divided into 2halves (yellow line) and each half was then incubated with stained witha Total Protein Stain (Fig. 6A; Revert 700 stain) which was detected inthe 700 nm channel and Anti-Phospho-EGF Receptor (EGFR) antibody(Fig. 6B) which was detected in the 800 nm channel. The portion of theblot with the blocked antibody cocktail (Anti-PhosphoEGFR + Phospho-EGFR immunogen peptide) has Revert 700 signal butno visible signal in the 800 nm channel (Fig. 6B; Right of the yellowline). Staining with a total protein stain ensures that equal protein wasloaded in all the lanes and it is not a lack of protein lysate that con-tributes to absent bands in the blocked portion of the membrane. Thisdata suggests that antibody against Phospho-EGFR protein detects aband ~175 kDa which is absent in the blocked blot. A separate blotcontaining equal amounts of protein extracts was divided into 2 halves(Fig. 6C and D) and multi-color detection of Pan-EGF (Fig. 6C) andPhospho-EGF (Fig. 6D) was performed. The portion of the blot with theblocked antibody cocktail (Anti-Phospho EGFR + Phospho-EGFR im-munogen peptide) has protein signal in the 700 nm channel from im-munodetection of Anti-Pan-EGFR (Fig. 6C) but no visible signal in the800 nm channel (Fig. 6D; Right of the yellow line).

4.1.2. SelectivityIn a typical quantitative analysis of the Western blot data, target

protein abundance may be lower relative to a large excess of unrelatedsample proteins. The antibody must be able to overcome this imbalanceand selectively bind the target antigen in a complex mixture, withoutinterference from off-target binding [90,92]. For this reason, selectivityshould be verified with endogenous levels of target expression in thecomplex sample [76,88]. Purified or overexpressed target protein altersthe balance of protein abundance in the sample, creating an artificialcontext that may not reflect actual antibody selectivity and expectedoff-target binding [88,105,106].

Optimization of certain assay conditions may improve antibodyselectivity. Insufficient antibody dilution and extended incubationtimes may promote off-target binding and detection of undesired,nonspecific bands [30,47,76,92]. Storage and reuse (“recycling”) ofdiluted primary antibodies is not recommended; this practice results ininconsistent antibody quality, titer, and stability that may introduceerror and make a quantitative analysis of Western blot data results lessreproducible [76]. Blocking buffer can also have a dramatic impact onantibody selectivity; an inappropriate blocking agent may greatly in-crease off-target binding in Western blot analysis [107,108].

4.1.3. Publication guidelinesDetailed reporting of antibody details and experimental methods is

an important aspect of quantitative analysis of Western blot data[53,87,109,110]. Recommendations for reporting of antibody in-formation are addressed elsewhere (including [76,87,110]). The au-thors should describe how antibody specificity was validated, includingthe methods used, assay parameters, positive and negative controls, andsample type (cell or tissue type, source, lysate, overexpressed or pur-ified target protein) The authors should be prepared to submit the rawvalidation data and unprocessed images during review [76,87,110].

4.2. Validate the internal loading control

An ILC must be present in all experimental samples at a stable level

Fig. 6. Validation of antibody against Phospho-EGFR in both untreatedand Etoposide treated A431 extracts via peptide blocking method.Untreated and Etoposide treated A431 cell extracts were loaded onto 4–12%Bis-Tris gel (Life Technologies NP0303; Lot# 16010670) and run under MOPSbuffer system. A single blot was split through the center (yellow line) with eachblot piece containing protein ladder. (A) After wet tank transfer and prior toblocking with a blocking buffer, the nitrocellulose blot was stained with Revert700 Total Protein Stain to visualize total protein extract loaded on each lane.The blot displays equal amounts of total protein loaded on each lane in the700 nm channel (red). (A′) The same blot in (A) was split into 2 pieces and thepiece on the left-hand side was incubated with Phospho-EGF Receptor (EGFR)(Tyr1068) (D7A5) XP® Rabbit mAb (CST #3777; Lot #13; 1 μg/ml) and anti-body binding was detected using IRDye® 800CW Goat (polyclonal) Anti-RabbitIgG (H + L) at 1/15000 dilution. The blot of the left was incubated with asolution of pre-incubated mixture of Rabbit Phospho-EGFR antibody andcustom synthesized blocking peptide (twice the molar ration of primary anti-body) corresponding to the antibody epitope of CST #3777 (CST #Y1068; Lot#5; 2 μg/ml). Both the blot pieces were incubated for one hour at room tem-perature with gentle shaking. Evident lack of signal in the blocked portion ofblot suggests that signal observed in the left blot is indeed specific to CST#3777. (B) CST #3777 antibody was further validated in A431 lysates byprobing the same blot with Mouse Anti-Human EGFR antibody cocktail(ThermoFisher Scientific AHR5062; Lot# 73759137A) and antibody bindingwas detected using IRDye® 800CW Goat (polyclonal) Anti-Mouse IgG (H+ L) at1/15000 dilution in the 700 nm channel to ensure that lanes have an equalamount of protein extracts as well stability of Pan-EGFR in the A431 extracts.Signal for Pan-EGFR detected by AHR5062 antibody is observed on both sidesof the blot suggesting that the blocking peptide is specific to CST #3777. (B′)The same blot in (A) was split into 2 pieces and the piece on the left-hand sidewas incubated with Phospho-EGF Receptor (EGFR) (Tyr1068) (D7A5) XP®Rabbit mAb (CST #3777; Lot #13; 1 μg/ml). The blot of the left was incubatedwith a solution of pre-incubated mixture of Rabbit Phospho-EGFR antibody andcustom synthesized blocking peptide (twice the molar ration of primary anti-body) corresponding to the antibody epitope of CST #3777 (CST #Y1068; Lot#5; 2 μg/ml). Both the blot pieces were incubated for one hour at room tem-perature with gentle shaking. Evident lack of signal in the blocked portion ofblot suggests that signal observed in the left blot is indeed specific to CST#3777. (For interpretation of the references to colour in this figure legend, thereader is referred to the web version of this article.)


7

Fig. 7. Identification and validation of an appropriate HKP in the MAPK/ERK pathway. Cell extracts prepared from untreated (UT) or Epidermal growth factor(EGF) (ET) treated A431 cells were loaded onto 10% 10% Bis-Tris gel (NP0303BOX) and run under the MOPS buffer system. All the blots were incubated withIntercept Blocking Buffer for one hour at room temperature. Technical replicates for both UT and ET categories are presented. Proteins belonging to MAPK/ERKpathway were selected to be tested via Western blot method. Total protein extract was visualized in each blot (a, a’, a”) using Revert 700 Total Protein Stain (LI-CORP/N 926-11010). Blots; b, b’, and b” were incubated with Anti-Ras antibody (CST #3965; Lot #1); Anti-CDK4 (D9G3E) antibody (CST #12790; Lot#14); and Anti-CDK6 (DCS83) antibody (CST #3136; Lot #4). Merged images of the blots (c, c’, c”) imaged under different acquisition channels; 700 nm for Revert 700 (a, a’, a”) and800 nm for Ras, CDK4, and CDK6 (b, b’, b”) highlight uniform loading of sample in each lane. Ras, CDK4, and CDK6 levels appear to be unchanged between untreatedand EGF treated A431 cell extracts (d, d’, d”). F test to compare variance, P value > 0.05). Sample loaded: 5 micrograms; Blocking buffer: Intercept Blocking Buffer(TBS); Intercept T20 (TBS) Antibody Diluent; Protein ladder: Chameleon duo ladder (LI-COR P/N 928-70000; Lot# C70803-03). Imager: Odyssey® CLx; resolution:169 μm; Intensity: auto mode.


8

to be used as a valid indicator or proxy of sample loading[3,29,33–35,37]. Stability should be demonstrated before an HKP orother single protein is used for quantitative analysis of Western blotdata normalization, to ensure that expression of the ILC is not affectedby experimental treatments or manipulations [29,33,34,47,53,111].

4.2.1. Stable or Unstable: ILC?Expression of the HKP or other single-protein control should be

carefully examined in samples that represent the desired experimentalconditions. A comparison of HKP levels to total protein loading in thesesamples is a straightforward way to demonstrate the stability of ex-pression. Alternatively, HKP levels may be compared to an unrelated,endogenous protein already shown to be stably expressed in the re-levant experimental samples and conditions. If changes are made inexperimental treatments, cell line, tissue type, cell density, or otherrelevant experimental parameters, stable HKP expression, as well as alinear range of detection, must be re-validated for the new conditions[33,34,47]. Several proteins belonging to the MAPK pathway weretested for stability in both vehicles treated and etoposide treated Jurkatcell extracts (Fig. 7c, c’, c”). Each blot was also stained to visualize total

protein loaded in each lane (Fig. 7a, a’, a”). No significant changes inprotein expression were observed for Ras, CDK4, and CDK6 proteins(Fig. 7b, b’, b”; d, d’, d”), and therefore, is suitable to be used as HKP forthe samples under consideration.

4.3. Two-sides of the same coin: Post-Translationally modified (PTM)proteins

Western blots generated to detect PTM proteins utilize an ILC thataccounts for the total presence of the unchanged protein (also known asPan-protein). Here, a modification-specific primary antibody is multi-plexed with a pan-specific primary antibody that recognizes the targetprotein in any modification state [6,38,71,72,112–114]. The pan- andmodification-specific antibodies should be derived from different hostspecies, so they can be discriminated by secondary antibodies labeledwith spectrally distinct fluorophores [63]. Pan/Phospho and othertypes of pan/PTM analysis are widely used to monitor and comparerelative changes in protein modification across a group of samples. Pan/Phospho analysis is specifically recommended by the Journal of Biolo-gical Chemistry, which stipulates that “phospho-specific antibody signals

Fig. 8. Determination of the combined linear range of detection for CPARP, β-ACTIN (ACTB), GAPDH, and COXIV. Jurkat cell extracts harvested from cellstreated with 25 μM Etoposide for 5 h at 37 °C was loaded on 4–12% Bis-Tris NuPage gels and electrophoresed under the MOPS buffer system. The blots were blockedwith Odyssey Blocking buffer- TBS for one hour at room temperature and subsequently incubated with primary antibodies against CPARP (CST #5625; Lot#13),ACTB (LI-COR P/N 926-42210; Lot #C80322-01), GAPDH (CST #5174; Lot #2), and COXIV (LI-COR P/N 926-42214; Lot #C5033324-02) (~1 μg/ml) at 4 °Covernight with gentle shaking. Subsequently, the primary antibody binding was visualized by using IRDye 800CW Goat anti-Rabbit IgG (H + L). Sample loaded:1.25–40 micrograms; Blocking buffer: Odyssey Blocking Buffer (TBS); Protein ladder: Chameleon duo ladder (LI-COR P/N 928-70000; Lot# C70803-03). Imager:Odyssey® CLx; resolution: 169 μm; Intensity: auto mode. Linear range of detection for all the proteins are plotted in (A). Linear range of CPARP and COXIV (B)appears to have overlapping range of detection compared to CPARP and ACTB (C) and CPARP and GAPDH (D).


9

should be normalized to total levels of the target protein” [97,111].Although pan/phospho analysis is the most commonly performed typeof pan/PTM analysis, other protein modifications are also studied withthis method (ubiquitination and palmitoylation analysis).

Pan-protein normalization accounts for changes in expression of thetarget protein, which might confound analysis of the abundance of themodified form. It also eliminates the stripping and reprobing of blots,which introduces error by causing loss of blotted proteins from themembrane [6,31,38]. Because the unmodified and modified targetprotein is detected on the same blot and in the same lane, normalizationcan correct for transfer artifacts in the blot. Pan/PTM analysis shouldonly be used for comparison of relative abundance; it does not provideinformation about the stoichiometry of the modification [77].

4.3.1. Publication guidelinesIf an HKP is used for quantitative analysis of Western blot data

normalization, authors should be prepared to provide evidence that itsexpression was not affected by the experimental treatments applied[53,111]. The editors of the Journal of Biological Chemistry have ex-pressed a preference for total protein normalization in quantitativeanalysis of Western blot data [53].

4.4. Define the linear range of detection

Using the validated antibodies, it is important to determine theappropriate amount of sample to load to generate data that can bequantified. Normalization and analysis must be performed in the linearrange of detection, where a linear and proportional response is ob-served between sample loading and band intensity.

4.4.1. Finding the combined linear range for a target protein and internalloading control

Careful examination of the linear range is particularly important forHKPs and other single-protein loading controls. The strong bands pro-duced by highly abundant ILC are frequently affected by saturation. Ifthe abundance of a target is substantially lower than the single-proteinloading control, the combined linear range may be very narrow. Totalprotein normalization typically provides a wider combined linearrange, without the rapid saturation observed for HKPs[35,44,50,51,55,56]. This wider combined linear range is very helpfulfor low-abundance target proteins, which often require the loading oflarge amounts of sample protein and cannot be used with a highlyabundant HKP.

The experiment in Fig. 8 demonstrates the differences in a linearrange that can be observed between different HKPs. Cleaved PARP (c-

Fig. 9. Combined Linear Range (CLR) detection of Total protein extract and OCT-4 in HEK293T cells. 4-fold dilution of cell extracts from HEK293T untreatedcontrol (a’-a”) as well as knockdown HEK293T (experimental) (b-b”) category were loaded onto 10% Bis-Tris gel (NP0303BOX) and run under the MOPS buffersystem. All the blots were incubated with Revert 700 Total Protein Stain (LI-COR P/N 926-11010) to visualize total protein extract in each lane. Subsequently, theblots were incubated in Intercept Blocking Buffer for one hour at room temperature and thereafter in Anti-Oct-4 antibody (CST #2750; Lot #4) and IRDye 800CWGoat anti-Rabbit IgG (H + L). (a-a”) Individual linear range of detection for total protein and OCT-4 was determined for control cell extracts (a - a”) as well as forexperimental cell extracts (b – b”). Combined linear range for total protein extract and OCT-4 protein detection was plotted for both cell extract types (c, d). CLR wasidentified at 4.5 micrograms for both cell types. Blocking buffer: Intercept Blocking Buffer (TBS); Intercept T20 (TBS) Antibody Diluent; Protein ladder: Chameleonduo ladder (LI-COR P/N 928-70000; Lot# C70803-03). Imager: Odyssey® CLx; resolution: 169 μm; Intensity: auto mode.


10

PARP) and three different HKPs (β-actin, COX IV, and GAPDH) weredetected in lysates from Jurkat cells treated with 25 μM Etoposide for5 h at 37 °C (check material and method section for more details).Because the target and HKPs span a range of MW and are well resolvedby electrophoresis, all four proteins could be detected simultaneouslyon a single Western blot. This facilitates comparison by eliminatingblot-to-blot inconsistencies in sample loading, electrophoresis, andtransfer. Fig. 8A shows a combined linear range for poly (ADP-ribose)polymerase cleavage (CPARP), GAPDH, COXIV, and β-actin (ACTB). InFig. 8B, the combined linear range for CPARP and COX IV is linearbetween ~7.5 and 30 micrograms. However, there is no useable com-bined linear range for CPARP, ACTB, and GAPDH (Fig. 8C and D). Inthese samples, band intensity for GAPDH and ACTB begins to saturateand plateau at ~15 μg of sample protein loading.

Working in the middle of the combined linear range helps to limiterror and variability introduced by high- and low-intensity data points.With an HKP or other single-protein loading control, high-intensity datapoints introduce error and should not be used for normalization; thesedata points will increase the mean Coefficient of Variance (CV) of thenormalized data and should be avoided [15,36]. From a statisticalperspective, variability introduced by a single-protein loading controlmay increase the frequency of false-negative results (small, but statis-tically significant, differences between samples that are not identifiedduring data analysis).

If validation experiments fail to identify a combined linear range ofdetection, as in Fig. 8C–D, a different ILC may be required. The use oftotal protein for normalization decreases the variability of high-in-tensity data points but does have the potential to increase variation forlow-intensity measurements [36]. Fig. 9 demonstrates the combinedlinear ranges for OCT-4 protein in HEK293T cell extracts derived fromcontrol and experimental categories. For both lysates, 4.5 microgramsappear to be in the middle of the combined linear range of detection forTotal protein extracts and OCT-4 protein. Optimization of transfermethods, sample loading, antibody choice, detection chemistry, orother assay parameters may help to improve the linear range of de-tection for the selected ILC. In general, increasing the number of in-ternal control proteins will reduce the mean CV and improve the ac-curacy of normalization, making it easier to reliably detect small orsubtle differences between samples [3,6,35,36,51].

4.4.2. Publication guidelinesWhen quantitative analysis of Western blot data results are re-

ported, authors should disclose how signal intensity was quantified aswell as how the linear relationship between sample loading and signalintensity was confirmed for each antigen [53,111].

4.5. Data analysis

Data analysis of Western blot data sets should be performed onbiological as well as technical replicates generated using the validatedantibodies, combined linear range of detection, and valid ILC estab-lished in the prior steps. After quantification of target and ILC signals,data analysis is performed.

Analysis typically begins with the designation of the experimentalcontrol sample that will be used for relative comparison. The lanenormalization factor is then calculated for each lane, using the HKPband intensity values or combined total protein signal values for eachlane (Fig. 10A). This factor is calculated by dividing the signal intensityvalue for each experimental sample by the signal intensity observed forthe control. To apply the normalization factors, the signal intensity ofthe target band in each lane is divided by the lane normalization factorfor that lane. This process generates the normalized signal intensityvalue for each sample.

=Fold changenormalized signal

normalized signalExperimental sample

Control sample

A ratiometric analysis is often performed to enable a relative com-parison of target abundance across a group of samples. The normalizedtarget signal for each sample is divided by the normalized target signalobserved in the control sample. These ratios express the abundance ofthe target protein as a fold or percentage change, relative to the control.A comparison of all samples to the experimental control produces re-lative values that are unitless, proportional, and independent of the rawsignal intensity. Fold change values > 1.0 indicate increased abun-dance relative to the control, and values < 1.0 denote decreasedabundance. Percentage change expresses the same information as apercentage, with a positive value indicating increased relative abun-dance and a negative percentage indicating decreased abundance(Table 1). Fig. 10 demonstrates the normalization of CPARP protein toTotal protein extracts from Jurkat cells exposed to three different ex-perimental conditions. Untreated, 33% spiked Etoposide lysate, and100% Etoposide treated lysates are loaded onto a PAGE gel in technicalreplicates (n = 3). Signal values for total protein, CPARP abundancefrom each category of cell extracts are represented in 10A′-B’. Fig. 10C’reveals the fold change in expression of CPARP in 33% (~2 fold) and100% (~3.3 fold) Etoposide exposed extracts.

4.5.1. Publication guidelinesMethods used to quantify and normalize signal intensities should be

disclosed, along with data analysis methods and software tools used forquantification and analysis. Raw data showing image analysis andquantification, including original digital images of Western blots, maybe requested during peer review [13,53,111,115].

4.6. Replicate and reproduce

Replication of quantitative analysis of Western blot data resultsconfirms the validity and reproducibility of any observed changes[13,14,18,20,21]. Replicate measurements help to ensure that the ex-perimental effects we report are reproducible – that they representactual differences between samples, rather than artifacts of experi-mental variability or noise [12,18]. Replication is essential when ana-lyzing small or subtle differences between samples, to characterize andunderstand the contribution of error and the limits of quantitativeanalysis [12,19,20]. Two types of replicates, technical and biological,are commonly used to address different questions in quantitative ana-lysis of Western blot data.

4.6.1. Technical and Biological ReplicatesTechnical replicates are repeated measurements of the same sample,

used to characterize the precision and variability of any assay ormethod [19,20,116]. Common examples include loading the samesample in multiple lanes on a Western blot, running replicate blots inparallel, or running the same kind of gel multiple times on differentdays. High variability between technical replicates makes it more dif-ficult to separate an observed experimental effect from the inherentvariation of the Q WB assay [12,19,20,53,117]. As a result, small butgenuine differences between samples may not be detected.

Biological replicates are parallel measurements of independent andbiologically distinct samples that are intended to control for randombiological variation [9,19–21]. A biologically relevant effect should beobserved reproducibly in independent samples. Analysis of samplesderived from multiple individuals or from multiple, independent cellcultures are common examples. In some cases, a similar experimentaleffect may be demonstrated in a different biological system or context[13,20,86,117]. Biological variability is generally expected to be higherthan technical variability [12,16,20]. Considerations for choosing ap-propriate replicates are described by others [19–21]. The replicationstrategy should be developed prior to the generation of quantitative


11

analysis of Western blot data [19,31] and it may be helpful to consult astatistician.

4.6.2. Using replicates to characterize error and noise in quantitativeanalysis of Western blot data

The coefficient of variation (CV) of replicate measurements can beused to evaluate the precision of quantitative analysis of Western blotdata results. A low CV indicates high precision of measurement and lowsignal variability in the assay; a larger CV denotes reduced precisionand greater assay variability. The CV can be used to determine if themagnitude of a relative change in target abundance is large enough tobe reliably detected above the assay noise. As a general rule of thumb,the magnitude of an observed effect should be at least twice as large asthe mean CV of the replicate measurements. In other words, for anobserved difference of 25% between samples (a 0.75-fold or 1.25-foldchange), the mean CV of replicate samples should be less than 12%. To

confidently report a small effect, it may be necessary to minimizesources of variation in the experimental protocol and increase thenumber of replicates to reduce the mean CV.

Normalization of Western blot data should always be performedbefore the analysis of individual replicate samples. Raw signal in-tensities should never be directly compared between blots. A multitudeof experimental factors influence the raw signal intensities observed oneach blot, and direct comparison of band intensity is meaningless. Rawband intensity is useful and informative only when analyzed by re-lative, ratiometric comparison to an appropriate control sample on thesame blot. To minimize error introduced by position effects, we re-commend loading replicate samples in random order for gel electro-phoresis. Consistent placement of certain samples in the same position(for example, always loading the positive control in the farthest leftlanes of the gel) may consistently produce edge effects that will beinadvertently propagated throughout the data analysis.

4.6.3. Publication guidelinesThe number of replicates performed for each quantitative analysis of

Western blot data measurement should be clearly indicated, with suf-ficient information to clearly distinguish technical replicates from in-dependent biological replicates [13,19,53,111]. Figure legends shouldaddress experimental uncertainty and explicitly define n for each ex-periment. Some journals also request that authors present a quantitativeanalysis of Western blot data and other small datasets as scatter plotswith error bars. This format more clearly depicts the spread and dis-tribution of data points than the traditional bar-and-plunger presenta-tion [53,109,118,119]. Fig. 10 illustrates how different data distribu-tions can produce the same bar graph – a graph that may suggest

Fig. 10. Normalization and analysisof PARP cleavage after treatmentwith etoposide. Jurkat cell extractsharvested from cells with the followingtreatments; untreated Jurkat lysate;33% (untreated lysate spiked with eto-poside-treated lysate at a final con-centration of 33%); 100% (etoposide-treated Jurkat lysate only). (treatedwith 25 μM Etoposide for 5 h at 37 °C)was loaded on 4–12% Bis-Tris NuPagegels and electrophoresed under theMOPS buffer system. A. Three blotswith biological replicates were in-cubated with Revert 700 Total ProteinStain (LI-COR P/N 926-11010) to vi-sualize total protein extract in eachlane. B–C. Subsequently, the blots wereblocked with Odyssey Blocking buffer-TBS for one hour at room temperatureand then incubated with primary anti-bodies against CPARP (CST #5625;Lot#13), Jurkat cells were treated withetoposide to induce cleavage of PARP.Cleaved PARP (c-PARP) was detectedin cell lysates by Western blotting(green). A′ Total protein staining (red)was used as an internal loading controlfor quantitative analysis. B′) Etoposide-induced cleavage of PARP was detectedon the membrane (graph shows rawdata and normalized result for eachtreatment). C′) Relative fold inductionof c-PARP was calculated for n = 3

biological replicates and n = 3 technical replicate per biological replicate, using the normalized values (graph shows individual fold-change values and mean).(n = 3 technical replicates for each sample; n = 3 biological replicates). CLR for total protein and CPARP was identified at 5 micrograms for all three cell types.Blocking buffer: Odyssey Blocking Buffer (TBS); Protein ladder: Chameleon duo ladder (LI-COR P/N 928-70000). Imager: Odyssey® CLx; resolution: 169 μm;Intensity: auto mode. Image Studio™ software was used for protein band detection. (For interpretation of the references to colour in this figure legend, the reader isreferred to the web version of this article.)

Table 1Abundance of target protein, expressed as fold or per-centage change relative to the control sample.

Fold Change Percentage change

1.0 0%1.25 +25%1.5 +50%2.0 +100%0.75 - 25%0.5 - 50%0.1 - 100%


12

different conclusions than the full dataset [118,119]. Weissgerber et al.[119] recently introduced free, online interactive tools designed toimprove data visualization for small datasets. The “interactive dot plot”tool allows different graphs to be viewed as univariate scatterplots, boxplots, and violin plots. It also enables visualization of clustered non-independent data, such as technical replicates. The “interactive re-peated experiments” is designed for visualization and comparison ofdata from repeated, independent experiments.

5. Discussion

Accurate, reproducible quantitative analysis of Western blot datarequires careful experimental design and execution. The core conceptsdescribed here should be considered during experimental design andplanning. Before a sample is lysed for protein extraction, careful con-sideration must be given to habitually disregarded aspects such as thecellular location of the target protein, purification strategy, as well ashow the extract will be quantified. Loading of the quantified sampleneeds further attention into the gel and buffer composition, and elec-trophoresis and protein transfer methods. Validation of antibodies usedfor immunodetection of the target protein should be undertaken by theuser and data shared with the readers. Identification of an appropriatehousekeeping protein that is stable is crucial and operating within thecombined linear range of the HKP and the target protein is even moreessential.

In recent years, standards and checklists have been proposed toenhance reproducibility by improving the transparency of reporting,increasing recognition of common experimental pitfalls, and raisingawareness about the importance of validation [13, 14, 109, 115,120–123]. These measures have stimulated conversation about theseimportant topics, which is critical for the development of the nomen-clature and resources needed to reach a consensus on guidelines andbest practice. Such guidelines must be a community effort and can behelpful to both new and established investigators in the field[13,14,121]. Standards and guidelines can also clarify and simplify thepeer review process. Han et al. [120] recently examined the impact of amanuscript submission checklist on the transparency and quality ofreporting in preclinical biomedical research. They evaluated the re-porting of methodological and analytical information in two journals:Nature (which implemented a mandatory checklist in 2013) and Cell(which does not require a checklist). Comparison of 2013 (pre-check-list) and 2015 (post-checklist) data shows that a required submissionchecklist is associated with improved reporting of methodological de-tails [120].

To be truly useful, guidelines cannot be overly rigid; in the la-boratory, one size does not fit all [13]. But flexible and appropriateguidelines can call attention to common mistakes, raise awareness, andmake researchers stop and think about the strengths and limitations ofthe methods they use. The systematic workflow we present here seeksto minimize the error introduced by common methodological limita-tions and improve the overall usefulness of quantitative analysis ofWestern blot data. We believe this flexible and adaptable approach hasthe potential to improve the quality, reproducibility, and precision ofquantitative analysis of Western blot data. It will also help to ensurethat the data generated in today's quantitative analysis of Western blotdata experiments will be able to meet increasingly rigorous publicationstandards in the future. The ultimate goal, of course, is to improve theoverall accuracy of quantitative analysis of Western blot data by in-creasing the likelihood that the changes we observe, and report arereproducible and reflect real and meaningful differences between bio-logical samples.

6. Materials and methods

6.1. Cell culture

A431 (ATCC® CRL-1555™), Jurkat (ATCC® TIB-152™) were culturedat 37 °C in a humidified atmosphere containing 5% CO2 as per ATCCguidelines. Cells were cultured up to 80% confluence before beingtreated or harvested to be stored at −80 °C.

6.2. Cell lysis

Cells were resuspended in radioimmunoprecipitation assay (RIPA)buffer (Pierce #89900) supplemented with Halt™ Protease andPhosphatase Inhibitor Cocktail (100×) (Thermo Scientific #878440)and incubated on ice for 20 min with intermittent vortexing. Lysateswere sonicated twice at 1-min intervals until no pellet was visible andthen clarified by centrifugation at 10,000 RCF for 15 min at 4 °C. Thesupernatant was taken and protein concentration determined by bi-cinchoninic acid (BCA) assay as previously described before beingstored at −80 °C.

6.3. Commercial cell lysates

C32 Whole Cell Lysate: sc-2205, POU5F1 overexpression lysate(Origene # LY400950; Lot #O1AF4D); HEK293T lysate (Origene#LY500001 Lot #O1AF4D); were mixed with either 2× protein loadingbuffer (PLB) (LICOR# 928-40004) OR 2× SDS buffer (OriGene) anddenatured by boiling at 97 °C for 5 min. The aliquots were then im-mediately placed on ice for 5 min before being spun down.

6.4. Western Blot

Lysates were loaded onto NuPAGE 4–12% 15 well Bis-Tris gels(Invitrogen; NP0323BOX), NuPAGE™ 10% Bis-Tris, 1.0 mm, 10-wellprotein gels (ThermoFisher Scientific; NP0301BOX) and run at 200 Vfor 37 min. Gels were wet transferred using 20% methanol onto eitherNitrocellulose membranes’ μM; 7 cm × 8.5 cm (P/N 926-31090).Membranes were blocked for one hour at room temperature in OdysseyBlocking Buffer (TBS) (LI-COR 927-50000). Primary antibodies werediluted with their respective blocking buffers (see figure legends) andincubated overnight at 4 °C. Washes were performed with TBS 0.1%Tween-20 (TBST) before the addition of secondary antibody for onehour at room temperature. Washes were performed with 1× TBSTbefore imaging on Odyssey CLx. protein detection was performed usingImage Studio Ver 5.2.

6.5. Antibodies

Primary antibodies include β-Actin (LI-COR 926-42212; Lot#C503324-02), CPARP (CST #5625; Lot#13), ACTB (LI-COR P/N 926-42210; Lot #C80322-01), GAPDH (CST #5174; Lot #2), and COXIV (LI-COR P/N 926-42214; Lot #C5033324-02), Phospho EGFR (CST #3777;Lot #13); Pan EGFR (Thermo Fisher Scientific AHR5062; Lot #737509137A), Ras (CST #3965; Lot #1), CDK4 (CST # 12790; Lot#14), CDK6 (CST #3136; Lot #4), OCT-4 (CST #2750; Lot #4).Primary antibodies were detected using IRDye 800CW Goat (poly-clonal) anti-Mouse IgG (H + L) highly cross-adsorbed (LI-COR#925-32210), IRDye 800CW Goat (polyclonal) anti-Rabbit IgG (H + L)highly cross-adsorbed (LI-COR# 926-32211).

6.6. Imaging

All blots were imaged wet on the Odyssey® CLx imaging systemusing 680 nm and 780 nm channels. Protein detection was performedusing Image Studio Ver 5.2. Adobe Photoshop Elements 13 was used toprepare image panels and annotations.


13

Author contributions

A.R.S.G co-wrote the main manuscript text and prepared Figures.J.A.H supervised, reviewed, and proof-read the manuscript. LPK de-signed and conducted the experiments, wrote the main manuscript textand finalized the Figures. ARSG and LPK are co-first authors of thismanuscript.

Declaration of competing interest

L. Pillai-Kastoori, A. Schutz-Geschwender, and Jeff A. Harford areemployees of LI-COR Biosciences.

Acknowledgments

We would like to sincerely thank Samuel Egel for critical review themanuscript; Jonathan Goodding for graphical contribution to themanuscript.

Appendix A. Supplementary data

Supplementary data related to this article can be found at https://doi.org/10.1016/j.ab.2020.113608.

References

[1] S.T. Towbin H, J. Gordon, Electrophoretic transfer of proteins from poly-acrylamide gels to nitrocellulose sheets: procedure and some applications, Proc.Natl. Acad. Sci. U. S. A. (1979) 76.

[2] W.N. Burnette, “Western Blotting”: electrophoretic transfer of proteins from so-dium dodecyl sulfate-polyacrylamide gels to unmodified nitrocellulose andradiographic detection with antibody and radioiodinated protein A, Anal.Biochem. 112 (1981) 195–203.

[3] G.M. Aldridge, D.M. Podrebarac, W.T. Greenough, I.J. Weiler, The use of totalprotein stains as loading controls: an alternative to high-abundance single-proteincontrols in semi-quantitative immunoblotting, J. Neurosci. Methods 172 (2008)250–254.

[4] S.L. Eaton, M.L. Hurtado, K.J. Oldknow, L.C. Graham, T.W. Marchant,T.H. Gillingwater, C. Farquharson, T.M. Wishart, A guide to modern quantitativefluorescent western blotting with troubleshooting strategies, JoVE 93 (2014),https://doi.org/10.3791/52099 e52099.

[5] P.M. Gerk, Quantitative immunofluorescent blotting of the multidrug resistance-associated protein 2 (MRP2), J. Pharmacol. Toxicol. Methods 63 (2011) 279–282.

[6] K.A. Janes, An analysis of critical factors for quantitative immunoblotting, Sci.Signal. 8 (2015) rs2.

[7] A.A. McDonough, L.C. Veiras, J.N. Minas, D.L. Ralph, Considerations whenquantitating protein abundance by immunoblot, Am. J. Physiol. Cell Physiol. 308(2015) C426–C433.

[8] L. Picariello, S. Carbonell Sala, V. Martineti, A. Gozzini, P. Aragona, I. Tognarini,M. Paglierani, G. Nesi, M.L. Brandi, F. Tonelli, A comparison of methods for theanalysis of low abundance proteins in desmoid tumor cells, Anal. Biochem. 354(2006) 205–212.

[9] N.R. Gough, Focus issue: tackling reproducibility and accuracy in cell signalingexperiments, Sci. Signal. 8 (2015) eg4.

[10] E. Marcus, Credibility and reproducibility, Canc. Cell 26 (2014) 771–772.[11] S.C. Taylor, T. Berkelman, G. Yadav, M. Hammond, A defined methodology for

reliable quantification of Western blot data, Mol. Biotechnol. 55 (2013) 217–226.[12] A. Casadevall, F.C. Fang, Reproducible science, Infect. Immun. 78 (2010)

4972–4975.[13] D.G. Drubin, Great science inspires us to tackle the issue of data reproducibility,

Mol. Biol. Cell 26 (2015) 3679–3680.[14] C.M. Lee, D.G. Drubin, All together now: how and why scientific communities

should develop best practice guidelines, Mol. Biol. Cell 27 (2016) 1707–1708.[15] J.P. Mollica, J.S. Oakhill, G.D. Lamb, R.M. Murphy, Are genuine changes in pro-

tein expression being overlooked? Reassessing Western blotting, Anal. Biochem.386 (2009) 270–275.

[16] A. Casadevall, F.C. Fang, Rigorous science: a how-to guide, MBio 7 (2016).[17] S.N. Goodman, D. Fanelli, J.P. Ioannidis, What does research reproducibility

mean? Sci. Transl. Med. 8 (2016) 341ps312.[18] A.L. Plant, L.E. Locascio, W.E. May, P.D. Gallagher, Improved reproducibility by

assuring confidence in measurements in biomedical research, Nat. Methods 11(2014) 895–898.

[19] P. Blainey, M. Krzywinski, N. Altman, Replication, Nat. Methods 11 (2014)879–880.

[20] K. Naegle, N.R. Gough, M.B. Yaffe, Criteria for biological reproducibility: whatdoes "n" mean? Sci. Signal. 8 (2015) fs7.

[21] D.L. Vaux, Know when your numbers are significant, Nature 492 (2012) 180–181.[22] D.L. Vaux, F. Fidler, G. Cumming, Replicates and repeats–what is the difference

and is it significant? A brief discussion of statistics and experimental design, EMBORep. 13 (2012) 291–296.

[23] J.J. Bass, D.J. Wilkinson, D. Rankin, B.E. Phillips, N.J. Szewczyk, K. Smith,P.J. Atherton, An overview of technical considerations for Western blotting ap-plications to physiological research, Scand. J. Med. Sci. Sports 27 (2017) 4–25.

[24] A. Pandey, K. Shin, R.E. Patterson, X.-Q. Liu, J.K. Rainey, Current strategies forprotein production and purification enabling membrane protein structuralbiology, Biochem. Cell. Biol. 94 (2016) 507–527.

[25] P.T. Wingfield, Overview of the purification of recombinant proteins, Curr. ProteinPept. Sci. 80 (2015) 6.1.1-6.1.35.

[26] M. Peach, N. Marsh, E.I. Miskiewicz, D.J. MacPhee, Solubilization of proteins: theimportance of Lysis buffer choice, in: B.T. Kurien, R.H. Scofield (Eds.), WesternBlotting: Methods and Protocols, Springer New York, New York, NY, 2015, pp.49–60.

[27] K. Gavini, K. Parameshwaran, Western Blot (Protein Immunoblot), StatPearls,(2019) Treasure Island (FL).

[28] E.I. Miskiewicz, D.J. MacPhee, Lysis buffer choices are key considerations to en-sure effective sample solubilization for protein electrophoresis, in: B.T. Kurien,R.H. Scofield (Eds.), Electrophoretic Separation of Proteins: Methods andProtocols, Springer New York, New York, NY, 2019, pp. 61–72.

[29] R.M. Murphy, G.D. Lamb, Important considerations for protein analyses usingantibody based techniques: down-sizing Western blotting up-sizes outcomes, J.Physiol. 591 (2013) 5823–5831.

[30] R. Ghosh, J.E. Gilda, A.V. Gomes, The necessity of and strategies for improvingconfidence in the accuracy of western blots, Expert Rev. Proteomics 11 (2014)549–560.

[31] A. Schutz-Geschwender, Western Blot Normalization: Challenges andConsiderations for Quantitative Analysis, LI-COR Biosciences, ResearchGate,2015.

[32] C.G. Perry, D.A. Kane, I.R. Lanza, P.D. Neufer, Methods for assessing mitochon-drial function in diabetes, Diabetes 62 (2013) 1041–1053.

[33] R. Baumgartner, E. Umlauf, M. Veitinger, S. Guterres, E. Rappold, R. Babeluk,G. Mitulovic, R. Oehler, M. Zellner, Identification and validation of platelet lowbiological variation proteins, superior to GAPDH, actin and tubulin, as tools inclinical proteomics, J Proteomics 94 (2013) 540–551.

[34] R.E. Ferguson, H.P. Carroll, A. Harris, E.R. Maher, P.J. Selby, R.E. Banks,Housekeeping proteins: a preliminary study illustrating some limitations as usefulreferences in protein expression studies, Proteomics 5 (2005) 566–571.

[35] Z.Z. Kirshner, R.B. Gibbs, Use of the REVERT((R)) total protein stain as a loadingcontrol demonstrates significant benefits over the use of housekeeping proteinswhen analyzing brain homogenates by Western blot: an analysis of samples re-presenting different gonadal hormone states, Mol. Cell. Endocrinol. 473 (2018)156–165.

[36] A. Degasperi, M.R. Birtwistle, N. Volinsky, J. Rauch, W. Kolch, B.N. Kholodenko,Evaluating strategies to normalise biological replicates of Western blot data, PLoSOne 9 (2014) e87293.

[37] R. Li, Y. Shen, An old method facing a new challenge: re-visiting housekeepingproteins as internal reference control for neuroscience research, Life Sci. 92 (2013)747–751.

[38] C.J. Bakkenist, R.K. Czambel, P.A. Hershberger, H. Tawbi, J.H. Beumer,J.C. Schmitz, A quasi-quantitative dual multiplexed immunoblot method to si-multaneously analyze ATM and H2AX Phosphorylation in human peripheral bloodmononuclear cells, Oncoscience 2 (2015) 542–554.

[39] C.J. Bakkenist, R.K. Czambel, Y. Lin, N.A. Yates, X. Zeng, J. Shogan, J.C. Schmitz,Quantitative analysis of ATM phosphorylation in lymphocytes, DNA Repair 80(2019) 1–7.

[40] X. Nie, C. Li, S. Hu, F. Xue, Y.J. Kang, W. Zhang, An appropriate loading control forwestern blot analysis in animal models of myocardial ischemic infarction, BiochemBiophys Rep 12 (2017) 108–113.

[41] S.L. Eaton, S.L. Roche, M. Llavero Hurtado, K.J. Oldknow, C. Farquharson,T.H. Gillingwater, T.M. Wishart, Total protein analysis as a reliable loading controlfor quantitative fluorescent Western blotting, PLoS One 8 (2013) e72457.

[42] X. Li, H. Bai, X. Wang, L. Li, Y. Cao, J. Wei, Y. Liu, L. Liu, X. Gong, L. Wu, S. Liu,G. Liu, Identification and validation of rice reference proteins for western blotting,J. Exp. Bot. 62 (2011) 4763–4772.

[43] V. Manceau, E. Kremmer, E.G. Nabel, A. Maucuer, The protein kinase KIS impactsgene expression during development and fear conditioning in adult mice, PLoSOne 7 (2012) e43946.

[44] C.P. Moritz, Tubulin or not tubulin: heading toward total protein staining asloading control in western blots, Proteomics 17 (2017).

[45] R. Perez-Perez, J.A. Lopez, E. Garcia-Santos, E. Camafeita, M. Gomez-Serrano,F.J. Ortega-Delgado, W. Ricart, J.M. Fernandez-Real, B. Peral, Uncovering suitablereference proteins for expression studies in human adipose tissue with relevance toobesity, PLoS One 7 (2012) e30326.

[46] J.G. Pinto, D.G. Jones, K.M. Murphy, Comparing development of synaptic proteinsin rat visual, somatosensory, and frontal cortex, Front. Neural Circ. 7 (2013) 97.

[47] S.D. Prokopec, J.D. Watson, R. Pohjanvirta, P.C. Boutros, Identification of re-ference proteins for Western blot analyses in mouse model systems of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) toxicity, PLoS One 9 (2014) e110730.

[48] M. Rocha-Martins, B. Njaine, M.S. Silveira, Avoiding pitfalls of internal controls:validation of reference genes for analysis by qRT-PCR and Western blotthroughout rat retinal development, PLoS One 7 (2012) e43028.

[49] S. Greer, R. Honeywell, M. Geletu, R. Arulanandam, L. Raptis, Housekeepinggenes; expression levels may change with density of cultured cells, J. Immunol.Methods 355 (2010) 76–79.

[50] M.A. Fortes, G.N. Marzuca-Nassr, K.F. Vitzel, C.H. da Justa Pinheiro,


14

https://doi.org/10.1016/j.ab.2020.113608

https://doi.org/10.1016/j.ab.2020.113608

http://refhub.elsevier.com/S0003-2697(19)30750-X/sref1











https://doi.org/10.3791/52099




























































































































P. Newsholme, R. Curi, Housekeeping proteins: how useful are they in skeletalmuscle diabetes studies and muscle hypertrophy models? Anal. Biochem. 504(2016) 38–40.

[51] J.S. Thacker, D.H. Yeung, W.R. Staines, J.G. Mielke, Total protein or high-abun-dance protein: which offers the best loading control for Western blotting? Anal.Biochem. 496 (2016) 76–78.

[52] T.A.J. Butler, J.W. Paul, E.-C. Chan, R. Smith, J.M. Tolosa, Misleading westerns:common quantification mistakes in western blot densitometry and proposed cor-rective measures, BioMed Res. Int. (2019) 5214821-5214821.

[53] A.J. Fosang, R.J. Colbran, Transparency is the key to quality, J. Biol. Chem. 290(2015) 29692–29694.

[54] C.J. Elbaggarari A, K. McDonald, A. Alburo, Evaluation of the Criterion Stain Free-TM Gel Imaging System for Use in Western Blotting Applications, Bio-RadLaboratories, 2008.

[55] C.P. Moritz, S.X. Marz, R. Reiss, T. Schulenborg, E. Friauf, Epicocconone staining:a powerful loading control for Western blots, Proteomics 14 (2014) 162–168.

[56] F. Heidebrecht, A. Heidebrecht, I. Schulz, S.E. Behrens, A. Bader, Improvedsemiquantitative Western blot technique with increased quantification range, J.Immunol. Methods 345 (2009) 40–48.

[57] M.A. Collins, J. An, D. Peller, R. Bowser, Total protein is an effective loadingcontrol for cerebrospinal fluid western blots, J. Neurosci. Methods 251 (2015)72–82.

[58] J.M. Dewe, B.L. Fuller, J.M. Lentini, S.M. Kellner, D. Fu, TRMT1-Catalyzed tRNAmodifications are required for redox homeostasis to ensure proper cellular pro-liferation and oxidative stress survival, Mol. Cell Biol. 37 (2017).

[59] R.C. Kuintzle, E.S. Chow, T.N. Westby, B.O. Gvakharia, J.M. Giebultowicz,D.A. Hendrix, Circadian deep sequencing reveals stress-response genes that adoptrobust rhythmic expression during aging, Nat. Commun. 8 (2017) 14529.

[60] K.M. Schubert, J. Qiu, S. Blodow, M. Wiedenmann, L.T. Lubomirov, G. Pfitzer,U. Pohl, H. Schneider, The AMP-related kinase (AMPK) induces Ca(2+)-in-dependent dilation of resistance arteries by interfering with actin filament for-mation, Circ. Res. 121 (2017) 149–161.

[61] D. Hawley, X. Tang, T. Zyrianova, M. Shah, S. Janga, A. Letourneau, M. Schicht,F. Paulsen, S. Hamm-Alvarez, H.P. Makarenkova, D. Zoukhri, Myoepithelial cell-driven acini contraction in response to oxytocin receptor stimulation is impairedin lacrimal glands of Sjogren's syndrome animal models, Sci. Rep. 8 (2018) 9919.

[62] J. Cackovic, S. Gutierrez-Luke, G.B. Call, A. Juba, S. O'Brien, C.H. Jun,L.M. Buhlman, Vulnerable Parkin loss-of-function Drosophila dopaminergic neu-rons have advanced mitochondrial aging, mitochondrial network loss and tran-siently reduced autophagosome recruitment, Front. Cell. Neurosci. 12 (2018) 39.

[63] A. Schutz-Geschwender, Y. Zhang, T. Holt, D. McDermitt, D.M. Olive,Quantitative, Two-Color Western Blot Detection with Infrared Fluorescence, LI-COR Biosciences, 2004.

[64] N. Plundrich, M.A. Lila, E. Foegeding, S. Laster, Protein-bound polyphenols create"ghost" band artifacts during chemiluminescence-based antigen detection,F1000Res 6 (2017) 254.

[65] K. Inoue, J. Rispoli, H. Kaphzan, E. Klann, E.I. Chen, J. Kim, M. Komatsu,A. Abeliovich, Macroautophagy deficiency mediates age-dependent neurodegen-eration through a phospho-tau pathway, Mol. Neurodegener. 7 (2012) 48.

[66] R.A.W. Stott, Enhanced chemiluminescence immunoassay, in: J.M. Walker (Ed.),The Protein Protocols Handbook, Humana Press, Totowa, NJ, 2002, pp.1089–1096.

[67] X. Wang, Y. Dong, A.J. Jiwani, Y. Zou, J. Pastor, M. Kuro-O, A.A. Habib, M. Ruan,D.A. Boothman, C.-R. Yang, Improved protein arrays for quantitative systemsanalysis of the dynamics of signaling pathway interactions, Proteome Sci. 9(2011) 53.

[68] T.P. Whitehead, L.J. Kricka, T.J. Carter, G.H. Thorpe, Analytical luminescence: itspotential in the clinical laboratory, Clin. Chem. 25 (1979) 1531.

[69] M. Zellner, R. Babeluk, M. Diestinger, P. Pirchegger, S. Skeledzic, R. Oehler,Fluorescence-based Western blotting for quantitation of protein biomarkers inclinical samples, Electrophoresis 29 (2008) 3621–3627.

[70] S.J. Charette, H. Lambert, P.J. Nadeau, J. Landry, Protein quantification by che-miluminescent Western blotting: elimination of the antibody factor by dilutionseries and calibration curve, J. Immunol. Methods 353 (2010) 148–150.

[71] H. Chen, J. Kovar, S. Sissons, K. Cox, W. Matter, F. Chadwell, P. Luan, C.J. Vlahos,A. Schutz-Geschwender, D.M. Olive, A cell-based immunocytochemical assay formonitoring kinase signaling pathways and drug efficacy, Anal. Biochem. 338(2005) 136–142.

[72] V. Boveia, A. Schutz-Geschwender, Quantitative analysis of signal transductionwith in-cell western immunofluorescence assays, Methods Mol. Biol. 1314 (2015)115–130.

[73] E. Bromage, L. Carpenter, S. Kaattari, M. Patterson, Quantification of coral heatshock proteins from individual coral polyps, Mar. Ecol. Prog. Ser. 376 (2009)123–132.

[74] Y.V. Wang, M. Wade, E. Wong, Y.C. Li, L.W. Rodewald, G.M. Wahl, Quantitativeanalyses reveal the importance of regulated Hdmx degradation for p53 activation,Proc. Natl. Acad. Sci. U. S. A. 104 (2007) 12365–12370.

[76] L. Pillai-Kastoori, S. Heaton, S.D. Shiflett, A.C. Roberts, A. Solache, A.R. Schutz-Geschwender, Antibody validation for western blot: by the user, for the user, J.Biol. Chem. 295 (4) (2019) 926–939, https://doi.org/10.1074/jbc.RA119.010472.

[77] K.A. Janes, An analysis of critical factors for quantitative immunoblotting, Sci.Signal. 8 (2015) rs2-rs2.

[78] J.J. Bass, D.J. Wilkinson, D. Rankin, B.E. Phillips, N.J. Szewczyk, K. Smith,P.J. Atherton, An overview of technical considerations for Western blotting ap-plications to physiological research, Scand. J. Med. Sci. Sports 27 (2017) 4–25.

[79] M. Gassmann, B. Grenacher, B. Rohde, J. Vogel, Quantifying Western blots: pitfallsof densitometry, Electrophoresis 30 (2009) 1845–1855.

[80] M. Berth, F. Michael Moser, M. Kolbe, J. Bernhardt, The State of the Art in theAnalysis of Two-Dimensional Gel Electrophoresis Images, (2007).

[81] I. Bio, -Rad Laboratories, Bio-Rad Laboratories, Inc. Image Lab™ Software Version6.0 Instrument Guide, Ver A., 2017.

[82] I. LI-COR Biosciences, LI-COR, Inc. Image Studio™ Software BackgroundSubtraction Guide, (2018).

[83] H.Y. Tan, T.W. Ng, Accurate step wedge calibration for densitometry of electro-phoresis gels, Optic Commun. 281 (2008) 3013–3017.

[84] F.T.a.W. Rasband, ImageJ User Guide, (2012).[85] I. LI-COR Biosciences, Empiria Studio™ Software: Adaptive Background

Subtraction, September (2018).[86] GBSI, The Science behind Antibody Validation Standards, Antibody Validation

Standards: strategy, Policies, Practices Workshop, GBSI, Asilomar, 2016.[87] GBSI, GBSI Report: Antibody Validation Standards, Policies, and Practices,

Antibody Validation Standards: Strategy, Policies, Practices Workshop, GBSI,Asilomar, 2016.

[88] M. Uhlen, A. Bandrowski, S. Carr, A. Edwards, J. Ellenberg, E. Lundberg,D.L. Rimm, H. Rodriguez, T. Hiltke, M. Snyder, T. Yamamoto, A proposal forvalidation of antibodies, Nat. Methods 13 (2016) 823–827.

[89] P. Acharya, A. Quinlan, V. Neumeister, The ABCs of finding a good antibody: howto find a good antibody, validate it, and publish meaningful data, F1000Res 6(2017) 851.

[90] U. Andreasson, A. Perret-Liaudet, L.J. van Waalwijk van Doorn, K. Blennow,D. Chiasserini, S. Engelborghs, T. Fladby, S. Genc, N. Kruse, H.B. Kuiperij, L. Kulic,P. Lewczuk, B. Mollenhauer, B. Mroczko, L. Parnetti, E. Vanmechelen,M.M. Verbeek, B. Winblad, H. Zetterberg, M. Koel-Simmelink, C.E. Teunissen, Apractical guide to immunoassay method validation, Front. Neurol. 6 (2015) 179.

[91] C.B. Saper, A guide to the perplexed on the specificity of antibodies, J. Histochem.Cytochem. 57 (2009) 1–5.

[92] M.G. Weller, Ten basic rules of antibody validation, Anal. Chem. Insights 13(2018) 1177390118757462.

[93] G. Roncador, P. Engel, L. Maestre, A.P. Anderson, J.L. Cordell, M.S. Cragg,V.C. Serbec, M. Jones, V.J. Lisnic, L. Kremer, D. Li, F. Koch-Nolte, N. Pascual,J.I. Rodriguez-Barbosa, R. Torensma, H. Turley, K. Pulford, A.H. Banham, TheEuropean antibody network's practical guide to finding and validating suitableantibodies for research, MAbs 8 (2016) 27–36.

[94] J. Bordeaux, A. Welsh, S. Agarwal, E. Killiam, M. Baquero, J. Hanna,V. Anagnostou, D. Rimm, Antibody validation, Biotechniques 48 (2010) 197–209.

[95] W. Olds, J. Li, siRNA knockdown validation 101: incorporating negative controlsin antibody research, F1000Res 5 (2016) 308.

[96] R.J. Colbran, Efforts to improve data transparency at the JBC, in: J.o.B. Chemistry(Ed.), Do's and Dont's of Data Analysis and Reporting, 2017 J Biol Chem..

[97] P. Acharya, A. Quinlan, V. Neumeister, The ABCs of finding a good antibody: howto find a good antibody, validate it, and publish meaningful data, F1000Res 6(2017) 851.

[98] L. Pillai-Kastoori, S. Heaton, S.D. Shiflett, A.C. Roberts, A. Solache, A.R. Schutz-Geschwender, Antibody validation for western blot: by the user, for the user, J.Biol. Chem. 295 (4) (2019) 926–939, https://doi.org/10.1074/jbc.RA119.010472.

[99] J. Bourbeillon, S. Orchard, I. Benhar, C. Borrebaeck, A. de Daruvar, S. Dubel,R. Frank, F. Gibson, D. Gloriam, N. Haslam, T. Hiltker, I. Humphrey-Smith,M. Hust, D. Juncker, M. Koegl, Z. Konthur, B. Korn, S. Krobitsch, S. Muyldermans,P.-A. Nygren, S. Palcy, B. Polic, H. Rodriguez, A. Sawyer, M. Schlapshy, M. Snyder,O. Stoevesandt, M.J. Taussig, M. Templin, M. Uhlen, S. van der Maarel,C. Wingren, H. Hermjakob, D. Sherman, Minimum information about a proteinaffinity reagent (MIAPAR), Nat. Biotechnol. 28 (2010) 650–653.

[100] GBSI, Approaches to Validation. The Science behind Antibody ValidationStandards, GBSI, 2016.

[101] S. Holmseth, Y. Dehnes, L.P. Bjørnsen, J.L. Boulland, D.N. Furness, D. Bergles,N.C. Danbolt, Specificity of antibodies: unexpected cross-reactivity of antibodiesdirected against the excitatory amino acid transporter 3 (EAAT3), Neuroscience136 (2005) 649–660.

[102] V. Kiermer, Antibodypedia, Nat. Methods 5 (2008) 860.[103] M. Uhlen, A. Bandrowski, S. Carr, A. Edwards, J. Ellenberg, E. Lundberg,

D.L. Rimm, H. Rodriguez, T. Hiltke, M. Snyder, T. Yamamoto, A Proposal forValidation of Antibodies, Nat Meth, advance online publication, 2016.

[104] F. Edfors, A. Hober, K. Linderbäck, G. Maddalo, A. Azimi, Å. Sivertsson, H. Tegel,S. Hober, C.A.-K. Szigyarto, L. Fagerberg, K. von Feilitzen, P. Oksvold, C. Lindskog,B. Forsström, M. Uhlen, Enhanced validation of antibodies for research applica-tions, Nat. Commun. 9 (2018) 4130.

[105] M. Signore, K.A. Reeder, Antibody validation by Western blotting, Methods Mol.Biol. 823 (2012) 139–155.

[106] M.C. Willingham, Conditional epitopes. is your antibody always specific? J.Histochem. Cytochem. 47 (1999) 1233–1236.

[107] K.L. Ambroz, Y. Zhang, A. Schutz-Geschwender, D.M. Olive, Blocking and detec-tion chemistries affect antibody performance on reverse phase protein arrays,Proteomics 8 (2008) 2379–2383.

[108] V. Kothari, S.T. Mathews, Detection of blotted proteins: not all blockers are cre-ated equal, Methods Mol. Biol. 1314 (2015) 27–32.

[109] Announcement, Towards greater reproducibility for life-sciences research inNature, Nature 546 (2017) 8.

[110] L.P. Freedman, J. Inglese, The increasing urgency for standards in basic biologicresearch, Canc. Res. 74 (2014) 4024–4029.

[111] Instructions for Authors, J. Biol. Chem. (2018), https://www.jbc.org/site/misc/


15














































































https://doi.org/10.1074/jbc.RA119.010472


























































































https://www.jbc.org/site/misc/ifora.xhtml

ifora.xhtml.[112] D. Bond, E. Foley, A quantitative RNAi screen for JNK modifiers identifies Pvr as a

novel regulator of Drosophila immune signaling, PLoS Pathog. 5 (2009) e1000655.[113] D. Bond, D.A. Primrose, E. Foley, Quantitative evaluation of signaling events in

Drosophila S2 cells, Biol. Proced. Online 10 (2008) 20–28.[114] N.T. Georgopoulos, L.A. Kirkwood, D.C. Walker, J. Southgate, Differential reg-

ulation of growth-promoting signalling pathways by E-cadherin, PLoS One 5(2010) e13621.

[115] J.E. Schwarzbauer, W.M. Leader, D.G. Drubin, Setting the bar for cell biology bestpractices, Mol. Biol. Cell 27 (2016) 2803.

[116] N.O.o.I. Research, NIH Rigor and Reproducibility Training Module 3: Biologicaland Technical Replicates, NIH Office of Intramural Research, NIH.

[117] A. Casadevall, L.M. Ellis, E.W. Davies, M. McFall-Ngai, F.C. Fang, A framework forimproving the quality of research in the biological sciences, MBio 7 (2016).

[118] T.L. Weissgerber, N.M. Milic, S.J. Winham, V.D. Garovic, Beyond bar and line

graphs: time for a new data presentation paradigm, PLoS Biol. 13 (2015)e1002128.

[119] T.L. Weissgerber, M. Savic, S.J. Winham, D. Stanisavljevic, V.D. Garovic,N.M. Milic, Data visualization, bar naked: a free tool for creating interactivegraphics, J. Biol. Chem. 292 (2017) 20592–20598.

[120] S. Han, T.F. Olonisakin, J.P. Pribis, J. Zupetic, J.H. Yoon, K.M. Holleran, K. Jeong,N. Shaikh, D.M. Rubio, J.S. Lee, A checklist is associated with increased quality ofreporting preclinical biomedical research: a systematic review, PLoS One 12(2017) e0183591.

[121] D.J. Klionsky, Developing a set of guidelines for your research field: a practicalapproach, Mol. Biol. Cell 27 (2016) 733–738.

[122] Reproducibility: let's get it right from the start, Nat. Commun. 9 (2018) 3716.[123] M. Mishra, S. Tiwari, A.V. Gomes, Protein purification and analysis: next gen-

eration Western blotting techniques, Expet Rev. Proteonomics 14 (2017)1037–1053.


16

https://www.jbc.org/site/misc/ifora.xhtml




























Documents

A systematic approach to quantitative Western blot analysis