Am J Physiol Cell Physiol 308: C426–C433, 2015. First published December 24, 2014; doi:10.1152/ajpcell.00400.2014.
Review
Considerations when quantitating protein abundance by immunoblot Alicia A. McDonough,1 Luciana C. Veiras,1 Jacqueline N. Minas,2 and Donna Lee Ralph1 1
Department of Cell and Neurobiology, Keck School of Medicine, University of Southern California, Los Angeles, California; and 2School of Natural Sciences, University of California, Merced, California Submitted 16 December 2014; accepted in final form 22 December 2014
McDonough AA, Veiras LC, Minas JN, Ralph DL. Considerations when quantitating protein abundance by immunoblot. Am J Physiol Cell Physiol 308: C426–C433, 2015. First published December 24, 2014; doi:10.1152/ajpcell.00400.2014.—The development of the immunoblot to detect and characterize a protein with an antisera, even in a crude mixture, was a breakthrough with wide-ranging and unpredictable applications across physiology and medicine. Initially, this technique was viewed as a tool for qualitative, not quantitative, analyses of proteins because of the high number of variables between sample preparation and detection with antibodies. Nonetheless, as the immunoblot method was streamlined and improved, investigators pushed it to quantitate protein abundance in unpurified samples as a function of treatment, genotype, or pathology. This short review, geared at investigators, reviewers, and critical readers, presents a set of issues that are of critical importance for quantitative analysis of protein abundance: 1) Consider whether tissue samples are of equivalent integrity and assess how handling between collection and assay influences the apparent relative abundance. 2) Establish the specificity of the antiserum for the protein of interest by providing clear images, molecular weight markers, positive and negative controls, and vendor details. 3) Provide convincing evidence for linearity of the detection system by assessing signal density as a function of sample loaded. 4) Recognize that loading control proteins are rarely in the same linear range of detection as the protein of interest; consider protein staining of the gel or blot. In summary, with careful attention to sample integrity, antibody specificity, linearity of the detection system, and acceptable loading controls, investigators can implement quantitative immunoblots to convincingly assess protein abundance in their samples. immunoblot; Western blot; antibody-antigen; immunodetection; quantitation
IN 1979, RENART, REISER, AND STARK
(22) reported a method to rapidly detect proteins as antigens after their resolution by native or SDS-PAGE. The procedure involved passive transfer of the proteins out of the gel onto paper coated with diazo groups that would covalently bind proteins as they were passively wicked out of the acrylamide gel. After blocking the remaining sites, the blot was incubated with diluted antibody, then with radiolabeled protein A, and finally analyzed by autoradiography. This method was a breakthrough for investigators because it not only provided a method to find a “needle in a haystack,” that is, a low-abundance protein in a homogenate, but also provided a method to efficiently screen a large number of antibodies, e.g., hybridoma samples, against a specific antigen. Although cumbersome (required coating paper with diazo groups and iodinating protein A), many labs used this method in the 1980s to address questions about protein subunit expression, assembly, and regulation; our own lab used these blots to study sodium pumps in a variety of tissues (16 –18, 24). Around the same time, other labs reported simpler procedures that involved electrophoretically blotting the gel onto nitrocellulose (5, 26), and working in the Pacific
Address for reprint requests and other correspondence: A. A. McDonough, Cell and Neurobiology, Keck School of Medicine of USC, 1333 San Pablo St. MMR 508, Los Angeles, CA 90033 (email: [email protected]). C426
Northwest, Burnette coined the term “Western blot,” in homage to Edwin Southern’s development of the DNA blot. Subsequent immunoblot simplifications included the use of secondary antibodies tagged with chemiluminescent or fluorescent probes to eliminate the need for radioactivity. The immunoblot is now, 35 years later, a ubiquitous tool used in most all aspects of biological research as well as a tool for commercial immunodiagnostics (4). While the Western blot was not originally designed to quantitate protein abundance with antibodies, this approach is quite commonly implemented to do just that. Our experience as investigators, reviewers, and readers has revealed a wide disparity in the rigor of execution and interpretation of immunoblot quantitation by authors and a similarly disparate rigor of reviewers’ attention to critical aspects of quantitation. The aim of this short review is to consider several issues that we have found to be important for immunoblot quantitation, including considering sample integrity and characteristics, demonstrating antibody specificity, establishing the linearity of the detection system, and implementing acceptable loading controls. Our report will not discuss how to optimize immunoblot signal-to-noise ratio, nor will we rank one detection and quantitation method versus another. There are many excellent resources available that describe how to perform and optimize immunoblots, including excellent vendor documents available online, reviews, and books (2, 12, 14). Our goal is to provide
0363-6143/15 Copyright © 2015 the American Physiological Society
http://www.ajpcell.org
Review C427
QUANTITATING IMMUNOBLOTS
an introduction to important considerations in quantitating immunoblot signals for both novices and experts as well as for reviewers and critical readers. We will use examples from tissues, but the same considerations apply to cell culture samples. Consider Sample Integrity and Properties Is the analysis comparing equivalent samples? Protein abundance is routinely normalized to the amount of protein loaded per lane (more on determining the actual amount to load is discussed below). If the experimental aim is to determine how a protein of interest changes with a treatment, disease process, genotype, or aging, it is important to consider confounding factors. Figure 1A provides images of normal and infarcted myocardium samples (at approximately the same magnification) to illustrate this point. About 50% of the cardiac muscle tissue is replaced by fibrotic tissue in the infarcted sample. Said another way, the total protein that will be used to normalize the relative abundance of the protein of interest is now made up of about 50/50 muscle and fibrotic tissue. Consequently, even if there is not a change in expression of the protein of interest in the muscle cells in the infarcted sample, its relative abundance will decrease about 50% when compared to normal myocardium. When studying highly fibrotic or otherwise modified tissues, the investigator must carefully consider the question under investigation and determine the best way to normalize the relative abundance of the protein of interest, for example, the abundance of a muscle-specific structural protein could be assessed in parallel by immunoblot and used to normalize the protein of interest. On the other hand, if the goal is to quantitate
A
the actual degree of fibrosis, then normalizing a fibrotic marker to a constant amount of sample protein could be appropriate. Obtaining relative abundance data usually requires accumulating a sufficient number of samples per group and, ideally, analyzing the groups on a single blot (with multiple gels on one blot paper if necessary) so that all the samples to be compared are incubated with the same aliquots of primary and secondary antibodies, and quantified simultaneously. These considerations reduce the variables that interfere with quantifying relative abundance in the groups. If all the samples cannot be collected and prepared on the same day, then the sample handling between collection and assay becomes an important variable that warrants serious consideration. Often tissues are quick frozen and then prepared for assay together. Figure 1B illustrates the results of a comparison of samples of mouse kidney homogenates prepared from kidneys quick frozen in liquid nitrogen and stored for a month at ⫺80°C, versus kidney homogenates prepared immediately after euthanizing the mice. Results are shown for the renal Na⫹/H⫹ exchanger isoform 3 (NHE3, the sodium transporter responsible for the bulk of the renal sodium reabsorption). The apparent abundance of NHE3 is less in the frozen versus the freshly prepared homogenates, indicating that there has been a partial loss of the epitope (perhaps by degradation or masking) as a consequence of freezing and thawing. The mouse samples, in our opinion, are still useful as long as they are prepared identically (all freshly prepared or all frozen for about the same time). However, our experience indicates that frozen rat kidneys do not fare as well as frozen mouse kidneys. Figure 1C illustrates samples from frozen and fresh rat kidneys prepared with the same reagents
B
C frozen
fresh
Normal myocardium
mw (kD)
mw (kD)
frozen
fresh
100
NHE3
100
NHE3
70 70
250
NKCC
55 100 70
Infarcted myocardium
250
NCC mouse kidney homogenate 45 µg/lane
100 70
rat kidney homogenate 40 µg/lane Fig. 1. Consider sample integrity. A: illustration showing that the percentage of normal cardiac myocytes can be significantly decreased in infarcted myocardium versus normal myocardium. B: illustration of the effect of freezing (versus not freezing) mouse kidneys before homogenization on the detection of the Na⫹/H⫹ exchanger isoform 3 (NHE3) epitope, a measure of the apparent abundance of NHE3. C: illustration of the near total loss of renal transporter epitope recognition when rat kidneys are frozen before preparation of homogenates. See text for details and discussion. NKCC, Na⫹-K⫹-2 Cl⫺ cotransporter; NCC, Na⫹-Cl⫺ cotransporter. AJP-Cell Physiol • doi:10.1152/ajpcell.00400.2014 • www.ajpcell.org
Review C428
QUANTITATING IMMUNOBLOTS
on the same day and analyzed on the same blot. While there are strong signals for the NHE3, the Na⫹-K⫹-2 Cl⫺ cotransporter (NKCC), and Na⫹-Cl⫺ cotransporter (NCC) in the freshly prepared homogenates, the signals are lost in kidneys that were freeze/thawed before homogenate preparation. For this reason, we recommend that the investigator always compare homogenates made from freshly isolated tissue to frozen tissue to understand if there is serious loss of antibody epitope for their protein of interest. Interestingly, we have found that once renal homogenates are prepared with protease and phosphatase inhibitors (19) and stored in single-use aliquots at ⫺80°C, the immunoblot signals of renal transporters do not appear to further decay over many years of storage, whether the samples were originally from fresh or freeze/thawed tissue (not shown). Many sample handling factors can affect relative abundance. As two examples, the phosphorylation status of the renal NCC is profoundly influenced by the time since consuming the last meal (25), and we have now been made aware of the influence of circadian rhythms on protein expression (23). In addition, sample enrichment by subcellular fractionation or affinity purification will exhibit batch-to-batch variability in relative recoveries; analysis of homogenates circumvents this recovery consideration. In summary, the investigator needs to consider the status of the tissues being compared and assess how handling between collection and assay can influence the calculation of a sample or treatment groups’ relative abundance.
Consider the Specificity of the Antibody for the Target It is the investigator’s responsibility to provide details about each antiserum sufficient for a reader to replicate the immunodetection. This requires providing not just the vendor information but also catalog number, because the vendor can have multiple antibodies to one protein. Information about antiserum dilution, host, and specific secondaries are also very useful because antibody vendors come and go while an antisera can be used by an author over decades. If many antisera are used, these can be organized in a table. Does the antiserum recognize the protein of interest? It is the investigator’s responsibility to answer this question. It goes without saying that every immunoblot needs to include molecular weight standards to assess and report the apparent molecular weight of the protein(s) detected by the antiserum. It is important to acknowledge that just because a vendor advertises that an antiserum recognizes a target protein does not guarantee that the antibody detects that protein in your samples, nor that it does not also bind to another unrelated protein (nonspecifically) with a similar molecular weight. As a case in point, a recent study by Herrera, and a companion commentary by the Eguchi lab, provide proof that seven commercially available antibodies to the angiotensin type 1 receptors (AT1R) lack specificity (7, 11). Bands of the “expected” apparent molecular weight were equally evident in tissues and cells that expressed no AT1R as in tissues known to express AT1R. The investiN-terminal antibody detects full length SPAK
Fig. 2. Consider the specificity of the antibody for the target. SPAK [sterile 20 (STE20)/SPS1-related proline/alanine-rich kinase] was detected first with an anti-NH2-terminal epitope antiserum, which detects full-length SPAK (FL-SPAK), then a red-tagged secondary; the blot was reprobed with an anti-COOH-terminal epitope antisera that detects FL-SPAK, smaller SPAK 2, and an even smaller kidney-specific SPAK isoforms (KS-SPAK), then a green-tagged secondary. Bottom: illustration of the image obtained when both signals were detected at the same time; yellow indicates detection at the same location. SPAK is expressed in testes as a single full-length isoform (FL-SPAK) included as a positive control, is absent in a SPAK knockout mouse (SPAK-KO), included as a negative control, and all three isoforms are present in wild-type mouse kidney. Nonspecific background bands (white arrows) were identified based on expression in the SPAK-KO. See text for details and discussion.
C-terminal antibody detects SPAK isoforms
SPAK double labeling FL-SPAK (yellow)
FL-SPAK (yellow) background
SPAK 2 (green) KS-SPAK (green)
background
KS-SPAK (green) testis
SPAK-KO
AJP-Cell Physiol • doi:10.1152/ajpcell.00400.2014 • www.ajpcell.org
WT
WT
Review C429
QUANTITATING IMMUNOBLOTS
phorylated form, and perhaps other bands that are in compartments that do not become phosphorylated, as we observed for renal NCC using nondenaturing gels (13). When using antiphospho-specific antisera, the main concern is that the antiserum may also cross react with the nonphosphorylated protein. To avoid this, it is recommended to preabsorb the antiserum with a nonphosphorylated version of the immunizing peptide and to keep the primary antiserum incubation time short to minimize cross-reaction with the nonphosphorylated protein. In summary, the investigator needs to convincingly establish the specificity of the antiserum for their protein of interest by assessing whether it is detected at the correct apparent molecular size (by providing a stained lane of molecular weight markers on the immunoblot); by determining whether it is present in a known positive control (often available for purchase from vendors); by determining whether it is absent in a sample that does not contain the protein of interest; and by identifying nonspecific “background” bands. Consider the Linearity of the Detection System How much protein should be assayed? Assuming that the integrity of the sample and specificity of the antiserum for the
A
2.5
α1
relative abundance (60 µg = 1)
gators postulate that this may be a common problem with G protein-coupled receptors and provide the opinion that authors (and vendors) must apply rigorous standards to establish the specificity of the antibody for the target. Given the significant effort required to characterize an antibody, as well as potential pitfalls, unbiased resources have begun to be established on the web to aid investigators in choosing antibodies that have a good chance of working based on previous published reports (e.g., www.Citeab.com) and epitope sequence considerations (e.g., www.uniprot.org). We provide a typical characterization for the transporteractivating kinase known as SPAK [sterile 20 (STE20)/SPS1related proline/alanine-rich kinase], using antibodies to COOH- and NH2-terminal epitopes, in Fig. 2. These antibodies, prepared by the Delpire lab for use in studying brain SPAK (20, 21), were applied by us to the study of kidney SPAK. Unlike brain SPAK, in the kidney, SPAK is expressed as three nested isoforms sharing a common COOH terminus: fulllength SPAK (FL-SPAK), SPAK 2, and a smaller kidneyspecific form (KS-SPAK) (15). The Delpire lab generously provided our lab with both antibodies as well as a kidney from a SPAK knockout (KO) mouse, to serve as a negative control. Initial characterization of the antibodies (20) revealed very strong SPAK signals in testis, which serves as a positive control. The top panel (Fig. 2) displays the immunoblot results obtained after probing with the anti-NH2-terminal epitope antibody, present only on FL-SPAK, followed by a secondary antibody with a red label: The testis sample lane gives a strong positive FL-SPAK signal, there is no staining in the SPAK KO kidney lane in the FL-SPAK region, and strong bands are evident at the position of FL-SPAK in the two wild-type (WT) mouse kidney sample lanes. However, there are three prominent bands below the FL-SPAK band coincident in both the WT and KO samples, indicating that these are nonspecific “background” bands. The middle panel displays the results obtained when the same blot was probed with an anti-COOHterminal epitope antiserum expected to detect all three of the nested renal SPAK isoforms, followed by a secondary antibody with a green label: A single strong band is evident in the positive control lane containing testis (which expresses only FL-SPAK), only a very faint background band is evident in the SPAK KO lane, indicating that this antisera is quite specific, and multiple bands are evident in the two WT lanes. The two images were superimposed, which indicates that 1) the FLSPAK, recognized by both the anti-COOH- and anti-NH2terminal antisera, appears yellow (indicating that they both recognize the same protein) in testis and the WT samples; 2) the SPAK 2 and KS-SPAK, which are recognized by only the COOH-terminal probe, appear green; and 3) “background bands,” detected by the NH2-terminal antibody, appear red. This example illustrates the utility of using fluorescent tagged secondary antibodies to multiple epitopes to discriminate between specific and non-specific labeling. With this information we were able to identify SPAK, SPAK 2 and KS-SPAK and determine how their levels were regulated by angiotensin II as well as nitric oxide synthase inhibition (8, 9). Fluorescent tagged secondaries are also useful to analyze protein phosphorylation because, if they are generated in two different species, one color can be used to detect the total protein and another for the phosphorylated form, and the overlay should detect a band labeled with both for the phos-
α2
2.0
α3 NCE
1.5
20
40
60
(µg)
80
100
1.0
0.5
0.0 0
20
40
60
80
100
protein loaded/lane (µg)
B
rat renal cortex
rat renal medulla
NKA α1 0.12
0.25
0.5
24
48
0.12 0.25
0.5 µg/lane
NKA β1 3
6
12
3
6
12
24
48 µg/lane
Fig. 3. Consider the linearity of the detection system. A: illustration of the immunodetection of sodium pump catalytic isoform subunits (␣1, ␣2, ␣3) and the Na⫹/Ca2⫹ exchanger (NCE) from human heart detected between 20 and 100 g/lane. [Redrawn from Wang et al. (27).] B: illustration of the immunodetection of sodium pump (NKA) ␣1-subunit between 0.12 and 0.5 g/lane and 1-subunit between 3 and 48 g/lane in rat kidney homogenates from cortex and medulla. See text for details and discussion.
AJP-Cell Physiol • doi:10.1152/ajpcell.00400.2014 • www.ajpcell.org
Review C430
QUANTITATING IMMUNOBLOTS
protein of interest are satisfactorily characterized, the investigator and critical reader should consider whether the conditions of sample loading and quantitation provide signals in the linear range of the detection system (regardless of whether the system is detecting light or radioactivity or fluorescence). That is, one should determine whether the signal generated from the primary and secondary antibody labeling is a direct measure of the relative abundance of the protein of interest in the samples. It is our opinion that linearity of the detection system should be established for each protein of interest in every study (ideally in every immunoblot) because of the high number of variables that can influence the final signal. Variables include the level of protein expression in the tissue, sample handling and preparation (Fig. 1), gel transfer conditions, blocking efficiency, antibody dilution, affinity for the target, and detection method (e.g., chemiluminescence emits a signal that changes with time); we contend that significant regulation is often missed or inaccurately assessed because immunoblots are analyzed without assessing linearity. We propose that investigators analyze multiple amounts of samples on the same blot to calibrate detection; Charette et al. (6) propose a similar strategy emanating from their analysis of nonlinear chemiluminescent signals. The inset of Fig. 3A provides a series of immunoblots of cardiac homogenates in which samples were loaded at 20, 40, 60, 80, and 100 g/lane then individually probed with antisera directed against sodium pump catalytic subunit isoforms (␣1, ␣2, and ␣3) or against the sodium/calcium exchanger (NCE) (redrawn from Ref. 27). The graph (Fig. 3A) plots relative abundance of the immunoblot signal from each protein of interest, normalized to the abundance of the signal from the 60 g sample defined as 1.0; the dotted line indicates equivalent
changes in protein loaded and density signal. Between 20 and 80 g/lane, the change in relative signal abundance is roughly equivalent to the amount of protein loaded for ␣2, ␣3, and NCE, but not for ␣1. This is a good illustration of a problematic issue because while the density of the ␣1 signal is “light,” compared with the NCE signal, the detection system is saturated above 60 g for ␣1, but not for NCE. Said another way, just because signals on an immunoblot are in the gray zone does not guarantee they are in a linear range of detection—a doubling of signal with abundance must be validated for the specific assay conditions. Figure 3B compares detection of renal sodium pump subunits (␣1 and 1). Despite the fact that the analysis shown is restricted to a linear range, it is clear that this linear range of detection for kidney ␣1 is attained with about 100-fold less homogenate than found for cardiac tissue, corresponding to the far greater abundance of Na-K-ATPase ␣1 in kidney versus heart. While this finding is not surprising, Fig. 3B also indicates that detection of the Na-K-ATPase 1 subunit in renal tissue is linear when loaded at 100-fold more than that for detection of ␣1. Why are the ranges not more similar for this ␣ heterodimer? One explanation is that 1 is a glycoprotein and the sugar residues reduce antibody-antigen binding; we and others established long ago that  detection improves when  is deglycosylated before immunodetection (3). Another explanation that needs to be considered for all immunodetection is that an antisera’s affinity for an antigen-containing protein is dependent on the biological reaction of the immunized host against the injected antigen during antisera production and, thus, is highly variable between hosts or hybridoma clones.
A 98.9
92.6
92.7
103.0
85.1
45.3
47.6
56.9
64.6
67.8
28.7
26.3
27.6
27.4
29.8
23.9
26.1
27.3
28.7
26.0
6µg/lane:
Fig. 4. Consider the loading controls: part 1. A: illustration of a single immunoblot constructed from two SDS-PAGE gels loaded with 6 g/lane (top of blot) and 3 g/lane (bottom part of blot) kidney homogenate and probed with antiserum to claudin 10 and a secondary antiserum tagged red. The quantitation, indicated to the right of the blot, demonstrates linearity of the detection system: Specifically, signal intensity increases 1.7-fold with sample doubling. B: image of the same blot reprobed with an anti-actin (pan-actin) antiserum and a secondary antiserum tagged green. The actin signal intensity does not change with sample doubling. We conclude that actin is not an acceptable loading control for this application; rather it is “overloaded” at 3 and 6 g kidney homogenate/lane. See text for further details and discussion.
3µg/lane:
B 6µg/lane:
3µg/lane:
AJP-Cell Physiol • doi:10.1152/ajpcell.00400.2014 • www.ajpcell.org
Review C431
QUANTITATING IMMUNOBLOTS
Consider the Loading Controls Is a loading control necessary if one has established the linearity of the detection system? Yes. Equivalent loading of sample amount per lane is important to establish if immunoblots will be quantitated. Unequal sample-to-sample loading can result from: high levels of variability in the determination of protein concentration of the samples, incorrect sample preparation, or poor pipetting into the wells. It is very common for investigators to reprobe their blots with what is known as a loading control or housekeeping protein (e.g., actin or GAPDH) that exhibits high-level, constitutive expression in the tissue and is not expected to vary with the treatment. However, a key problem with this approach is that these loading control proteins are rarely in the same linear range of detection as the target protein of interest. A number of studies concur that in the typical loading range of 10 –50 g of cell lysate, quantitation of housekeeping proteins by immunodetection is not possible because they are in saturating quantities (1, 10, 28). A good illustration of the issue is provided in Fig. 4 in which five samples of renal homogenates were resolved at 6 and 3 g/lane (to establish linearity). The immunoblot was first probed with an antiserum directed to claudin 10 (a tight junction protein) and a red-tagged secondary (Fig. 4A, left) and claudin 10 was detected between the 17 and 28 kDa molecular mass markers. Quantitation of the five samples, provided to the right, indicates a good linearity of the detection system as the signals were roughly doubled with 6 versus 3 g protein loading. The same blot was probed with an antiserum to actin and a green-tagged secondary, and actin was detected above 34 kDa (Fig. 4B). The signal intensity of the actin band does not significantly change when sample loaded was reduced to ½; thus, actin is not an acceptable loading control for this application; it is saturating the membrane at just 3 g/lane. An alternative to a single loading control is protein staining of the gel or blot as a loading control (1, 28). Vendors are starting to market kits to stain loading gels, but it can be done
with what is on hand in the lab. We make up enough of each sample to run a parallel “loading gel” which is loaded and run identically to the gel(s) that will be immunoblotted (another approach is to actually stain the blot but we are concerned that this could interfere with immunodetection). In our approach, used by others as well, the gel is not blotted, rather, it is simply stained and destained and a number of bands are chosen to quantitate by densitometry. Another approach is to quantitate the density of the entire lane from top to bottom, but we have had good results assessing discrete bands that do not appear to be too dark or too light. Figure 5 provides an illustration of a loading gel containing eight samples that were all, nominally, loaded with 7.5 g/lane of renal cortex homogenate. Five bands were identified, quantitated by densitometry, then normalized to the mean density for that band in all eight samples, defined as 1.0. The relative abundance results for each sample as well as the average value for the five bands within each sample are displayed below the loading gel (Fig. 5). A visual inspection of the loading gel suggests samples 4 and 7 are relatively overloaded and the quantitation bears this out, as both have relative average densities of 1.13; the others vary by 5% or less. At this point, the investigator can report the level of variability that is acceptable for the quantitative immuno-
1
2 3
4 5
Loading gel- 7.5 µg renal cortex/lane 1.2
Relative Density
In summary, if an investigator aims to quantify changes in protein expression (or phosphorylation) using immunoblots, it is our experience and opinion that each immunoblot should contain a demonstration of the linearity of the detection system. This can be accomplished in a number of ways. We routinely run samples at 1 and ½ protein amounts (which may require multiple gels that are transferred together on the same blot membrane) and then process the single blot through immunodetection using the identical primary and secondary antibodies and simultaneous quantitation (as in Fig. 4). We calculate the linearity ratio by dividing the signal obtained with loading of the full amount with signal obtained with ½ the amount and expect it to be close to 2. If not, then the assay is underestimating or overestimating the effect of the treatment. In some cases we have included blots containing both 1 and ½ loading amounts in our figures (19) and in others we have reported the method used to establish linearity and shown just one amount in the figures (9). In this latter case, we have been asked by reviewers to provide images of every immunoblot loaded at the two amounts for verification. Another option is to provide a linearity analysis (as in Fig. 3A) as a supplemental figure, if the journal allows.
Band 1 1.1
Band 2 Band 3
1.0
Band 4 Band 5
0.9
AVERAGE 0.8 0
1
2
4
3
5
6
7
8
sample #
Fig. 5. Consider the loading controls: part 2. Direct protein staining of the gel or blot is an alternative to the loading control to assess whether the same amount of protein is loaded per lane. Top: image of a stained post SDS-PAGE gel (not blot) of renal homogenate samples from eight rat kidneys loaded at 7.5 g/lane. Arrows indicate the 5 unidentified protein bands arbitrarily chosen for quantitation to determine whether the 8 samples were equivalently loaded. Values were normalized to the mean value of each band in the 8 samples defined as 1.0. Bottom: illustration of the relative density of each of the 5 arbitrary bands (color coded) in each sample as well as the average of the 5 bands (black). The results indicate that samples 1, 2, 3, 6, and 8 are very close to equivalent with average relative density of 1.0 while 4, 5, and 7 deviate, by ⬍13%. See text for further details and discussion.
AJP-Cell Physiol • doi:10.1152/ajpcell.00400.2014 • www.ajpcell.org
Review C432
QUANTITATING IMMUNOBLOTS
blots, and options for proceeding include correcting the density of the immunoblots for this variance, beginning again with the protein assays and sample preparation to reduce sample to sample variability to 5–10% from the mean average density or less, or reporting and accepting the variability after providing a rationale. In summary, staining the gel provides a direct image of the actual protein loaded. The stained gel approach is a far more valuable loading control than that obtained by quantitating saturated signals from housekeeping proteins. The latter reveals little more than whether the sample was loaded or not. Conclusion During the past 35 years, immunoblotting has been a critical tool for making discoveries and progress in many areas of investigation from early use in expression cloning of genes, to utilization for studies of protein expression, to screening antibodies and clinical immunodiagnostics. Physiologists, with their focus on how tissues and organ systems work normally and in response to challenges, have relied on the Western blot’s ability to detect a “needle in a haystack” to address questions about organ system operation and regulation. However, since the method is simple to set up and implement, results are often conducted and reported without the requisite controls. Samples and reagents must be validated and controlled to obtain interpretable results. Said another way, if an investigator aims to determine whether there are more needles in one haystack than another, it is key to convincingly validate the integrity of the sample, the specificity of the antibody, and the linearity of the detection system and to assess sample loading. NOTE ADDED IN PROOF We recently became aware of a very useful review by Murphy and Lamb (18a) that similarly reviews important considerations for quantitating proteins in muscle using antibody-based techniques. ACKNOWLEDGMENTS Drs. Jo Adams and Christina Bennett provided very helpful comments. GRANTS This work was supported by National Institute of Diabetes and Digestive and Kidney Diseases Grant DK-083785 (to A. A. McDonough). DISCLOSURES No conflicts of interest, financial or otherwise, are declared by the author(s). AUTHOR CONTRIBUTIONS A.A.M., L.C.V., J.N.M., and D.L.R. performed experiments; A.A.M. prepared figures; A.A.M. drafted manuscript; A.A.M., L.C.V., J.N.M., and D.L.R. edited and revised manuscript; A.A.M., L.C.V., J.N.M., and D.L.R. approved final version of manuscript. REFERENCES 1. Aldridge GM, Podrebarac DM, Greenough WT, Weiler IJ. The use of total protein stains as loading controls: an alternative to high-abundance single-protein controls in semi-quantitative immunoblotting. J Neurosci Methods 172: 250 –254, 2008. 2. Alegria-Schaffer A, Lodge A, Vattem K. Performing and optimizing Western blots with an emphasis on chemiluminescent detection. Methods Enzymol 463: 573–599, 2009. 3. Azuma KK, Hensley CB, Tang MJ, McDonough AA. Thyroid hormone specifically regulates skeletal muscle Na⫹-K⫹-ATPase ␣2- and 2-isoforms. Am J Physiol Cell Physiol 265: C680 –C687, 1993.
4. Burnette WN. Western blotting: remembrance of past things. Methods Mol Biol 536: 5–8, 2009. 5. Burnette WN. “Western blotting”: electrophoretic transfer of proteins from sodium dodecyl sulfate–polyacrylamide gels to unmodified nitrocellulose and radiographic detection with antibody and radioiodinated protein A. Anal Biochem 112: 195–203, 1981. 6. Charette SJ, Lambert H, Nadeau PJ, Landry J. Protein quantification by chemiluminescent Western blotting: elimination of the antibody factor by dilution series and calibration curve. J Immunol Methods 353: 148 –150, 2010. 7. Elliott KJ, Kimura K, Eguchi S. Lack of specificity of commercial antibodies leads to misidentification of angiotensin type-1 receptor protein. Hypertension 61: e31, 2013. 8. Giani JF, Janjulia T, Kamat N, Seth DM, Blackwell WL, Shah KH, Shen XZ, Fuchs S, Delpire E, Toblli JE, Bernstein KE, McDonough AA, Gonzalez-Villalobos RA. Renal angiotensin-converting enzyme is essential for the hypertension induced by nitric oxide synthesis inhibition. J Am Soc Nephrol 25: 2752–2763, 2014. 9. Gonzalez-Villalobos RA, Janjoulia T, Fletcher NK, Giani JF, Nguyen MT, Riquier-Brison AD, Seth DM, Fuchs S, Eladari D, Picard N, Bachmann S, Delpire E, Peti-Peterdi J, Navar LG, Bernstein KE, McDonough AA. The absence of intrarenal ACE protects against hypertension. J Clin Invest 123: 2011–2023, 2013. 10. Hammond M, Kohn J, Oh K, Piatti P, Liu N. A method for greater reliability in Western blot loading controls: stain-free total protein quantitation. Bio-Rad Bulletin 6360, 2013. 11. Herrera M, Sparks MA, Alfonso-Pecchio AR, Harrison-Bernard LM, Coffman TM. Lack of specificity of commercial antibodies leads to misidentification of angiotensin type 1 receptor protein. Hypertension 61: 253–258, 2013. 12. Kurien BT, Scofield Hal R. Protein Blotting and Detection. Methods Mol Biol 536: 2009. 13. Lee DH, Maunsbach AB, Riquier-Brison AD, Nguyen MT, Fenton RA, Bachmann S, Yu AS, McDonough AA. Effects of ACE inhibition and ANG II stimulation on renal Na-Cl cotransporter distribution, phosphorylation, and membrane complex properties. Am J Physiol Cell Physiol 304: C147–C163, 2013. 14. Mahmood T, Yang PC. Western blot: technique, theory, and trouble shooting. N Am J Med Sci 4: 429 –434, 2012. 15. McCormick JA, Mutig K, Nelson JH, Saritas T, Hoorn EJ, Yang CL, Rogers S, Curry J, Delpire E, Bachmann S, Ellison DH. A SPAK isoform switch modulates renal salt transport and blood pressure. Cell Metab 14: 352–364, 2011. 16. McDonough A. Immunodetection of Na,K-ATPase in guinea-pig retinal layers, cornea and lens. Exp Eye Res 40: 667–674, 1985. 17. McDonough A, Schmitt C. Comparison of subunits of cardiac, brain, and kidney Na⫹-K⫹-ATPase. Am J Physiol Cell Physiol 248: C247– C251, 1985. 18. McDonough AA, Hiatt A, Edelman IS. Characteristics of antibodies to guinea pig (Na⫹ ⫹ K⫹)-adenosine triphosphatase and their use in cell-free synthesis studies. J Membr Biol 69: 13–22, 1982. 18a.Murphy RM, Lamb GD. Important considerations for protein analyses using antibody based techniques: down-sizing Western blotting up-sizes outcomes. J Physiol 591: 5823–5831, 2013. 19. Nguyen MT, Yang LE, Fletcher NK, Lee DH, Kocinsky H, Bachmann S, Delpire E, McDonough AA. Effects of K⫹-deficient diets with and without NaCl supplementation on Na⫹, K⫹, and H2O transporters’ abundance along the nephron. Am J Physiol Renal Physiol 303: F92–F104, 2012. 20. Piechotta K, Garbarini N, England R, Delpire E. Characterization of the interaction of the stress kinase SPAK with the Na⫹-K⫹-2Cl⫺ cotransporter in the nervous system: evidence for a scaffolding role of the kinase. J Biol Chem 278: 52848 –52856, 2003. 21. Piechotta K, Lu J, Delpire E. Cation chloride cotransporters interact with the stress-related kinases Ste20-related proline-alanine-rich kinase (SPAK) and oxidative stress response 1 (OSR1). J Biol Chem 277: 50812–50819, 2002. 22. Renart J, Reiser J, Stark GR. Transfer of proteins from gels to diazobenzyloxymethyl-paper and detection with antisera: a method for studying antibody specificity and antigen structure. Proc Natl Acad Sci USA 76: 3116 –3120, 1979.
AJP-Cell Physiol • doi:10.1152/ajpcell.00400.2014 • www.ajpcell.org
Review QUANTITATING IMMUNOBLOTS 23. Richards J, Gumz ML. Mechanism of the circadian clock in physiology. Am J Physiol Regul Integr Comp Physiol 304: R1053–R1064, 2013. 24. Schmitt CA, McDonough AA. Developmental and thyroid hormone regulation of two molecular forms of Na⫹-K⫹-ATPase in brain. J Biol Chem 261: 10439 –10444, 1986. 25. Sorensen MV, Grossmann S, Roesinger M, Gresko N, Todkar AP, Barmettler G, Ziegler U, Odermatt A, Loffing-Cueni D, Loffing J. Rapid dephosphorylation of the renal sodium chloride cotransporter in response to oral potassium intake in mice. Kidney Int 83: 811–824, 2013.
C433
26. Towbin H, Staehelin T, Gordon J. Electrophoretic transfer of proteins from polyacrylamide gels to nitrocellulose sheets: procedure and some applications. Proc Natl Acad Sci USA 76: 4350 –4354, 1979. 27. Wang J, Schwinger RH, Frank K, Muller-Ehmsen J, Martin-Vasallo P, Pressley TA, Xiang A, Erdmann E, McDonough AA. Regional expression of sodium pump subunits isoforms and Na⫹-Ca⫹⫹ exchanger in the human heart. J Clin Invest 98: 1650 –1658, 1996. 28. Welinder C, Ekblad L. Coomassie staining as loading control in Western blot analysis. J Proteome Res 10: 1416 –1419, 2011.
AJP-Cell Physiol • doi:10.1152/ajpcell.00400.2014 • www.ajpcell.org
Review
Considerations when quantitating protein abundance by immunoblot Alicia A. McDonough,1 Luciana C. Veiras,1 Jacqueline N. Minas,2 and Donna Lee Ralph1 1
Department of Cell and Neurobiology, Keck School of Medicine, University of Southern California, Los Angeles, California; and 2School of Natural Sciences, University of California, Merced, California Submitted 16 December 2014; accepted in final form 22 December 2014
McDonough AA, Veiras LC, Minas JN, Ralph DL. Considerations when quantitating protein abundance by immunoblot. Am J Physiol Cell Physiol 308: C426–C433, 2015. First published December 24, 2014; doi:10.1152/ajpcell.00400.2014.—The development of the immunoblot to detect and characterize a protein with an antisera, even in a crude mixture, was a breakthrough with wide-ranging and unpredictable applications across physiology and medicine. Initially, this technique was viewed as a tool for qualitative, not quantitative, analyses of proteins because of the high number of variables between sample preparation and detection with antibodies. Nonetheless, as the immunoblot method was streamlined and improved, investigators pushed it to quantitate protein abundance in unpurified samples as a function of treatment, genotype, or pathology. This short review, geared at investigators, reviewers, and critical readers, presents a set of issues that are of critical importance for quantitative analysis of protein abundance: 1) Consider whether tissue samples are of equivalent integrity and assess how handling between collection and assay influences the apparent relative abundance. 2) Establish the specificity of the antiserum for the protein of interest by providing clear images, molecular weight markers, positive and negative controls, and vendor details. 3) Provide convincing evidence for linearity of the detection system by assessing signal density as a function of sample loaded. 4) Recognize that loading control proteins are rarely in the same linear range of detection as the protein of interest; consider protein staining of the gel or blot. In summary, with careful attention to sample integrity, antibody specificity, linearity of the detection system, and acceptable loading controls, investigators can implement quantitative immunoblots to convincingly assess protein abundance in their samples. immunoblot; Western blot; antibody-antigen; immunodetection; quantitation
IN 1979, RENART, REISER, AND STARK
(22) reported a method to rapidly detect proteins as antigens after their resolution by native or SDS-PAGE. The procedure involved passive transfer of the proteins out of the gel onto paper coated with diazo groups that would covalently bind proteins as they were passively wicked out of the acrylamide gel. After blocking the remaining sites, the blot was incubated with diluted antibody, then with radiolabeled protein A, and finally analyzed by autoradiography. This method was a breakthrough for investigators because it not only provided a method to find a “needle in a haystack,” that is, a low-abundance protein in a homogenate, but also provided a method to efficiently screen a large number of antibodies, e.g., hybridoma samples, against a specific antigen. Although cumbersome (required coating paper with diazo groups and iodinating protein A), many labs used this method in the 1980s to address questions about protein subunit expression, assembly, and regulation; our own lab used these blots to study sodium pumps in a variety of tissues (16 –18, 24). Around the same time, other labs reported simpler procedures that involved electrophoretically blotting the gel onto nitrocellulose (5, 26), and working in the Pacific
Address for reprint requests and other correspondence: A. A. McDonough, Cell and Neurobiology, Keck School of Medicine of USC, 1333 San Pablo St. MMR 508, Los Angeles, CA 90033 (email: [email protected]). C426
Northwest, Burnette coined the term “Western blot,” in homage to Edwin Southern’s development of the DNA blot. Subsequent immunoblot simplifications included the use of secondary antibodies tagged with chemiluminescent or fluorescent probes to eliminate the need for radioactivity. The immunoblot is now, 35 years later, a ubiquitous tool used in most all aspects of biological research as well as a tool for commercial immunodiagnostics (4). While the Western blot was not originally designed to quantitate protein abundance with antibodies, this approach is quite commonly implemented to do just that. Our experience as investigators, reviewers, and readers has revealed a wide disparity in the rigor of execution and interpretation of immunoblot quantitation by authors and a similarly disparate rigor of reviewers’ attention to critical aspects of quantitation. The aim of this short review is to consider several issues that we have found to be important for immunoblot quantitation, including considering sample integrity and characteristics, demonstrating antibody specificity, establishing the linearity of the detection system, and implementing acceptable loading controls. Our report will not discuss how to optimize immunoblot signal-to-noise ratio, nor will we rank one detection and quantitation method versus another. There are many excellent resources available that describe how to perform and optimize immunoblots, including excellent vendor documents available online, reviews, and books (2, 12, 14). Our goal is to provide
0363-6143/15 Copyright © 2015 the American Physiological Society
http://www.ajpcell.org
Review C427
QUANTITATING IMMUNOBLOTS
an introduction to important considerations in quantitating immunoblot signals for both novices and experts as well as for reviewers and critical readers. We will use examples from tissues, but the same considerations apply to cell culture samples. Consider Sample Integrity and Properties Is the analysis comparing equivalent samples? Protein abundance is routinely normalized to the amount of protein loaded per lane (more on determining the actual amount to load is discussed below). If the experimental aim is to determine how a protein of interest changes with a treatment, disease process, genotype, or aging, it is important to consider confounding factors. Figure 1A provides images of normal and infarcted myocardium samples (at approximately the same magnification) to illustrate this point. About 50% of the cardiac muscle tissue is replaced by fibrotic tissue in the infarcted sample. Said another way, the total protein that will be used to normalize the relative abundance of the protein of interest is now made up of about 50/50 muscle and fibrotic tissue. Consequently, even if there is not a change in expression of the protein of interest in the muscle cells in the infarcted sample, its relative abundance will decrease about 50% when compared to normal myocardium. When studying highly fibrotic or otherwise modified tissues, the investigator must carefully consider the question under investigation and determine the best way to normalize the relative abundance of the protein of interest, for example, the abundance of a muscle-specific structural protein could be assessed in parallel by immunoblot and used to normalize the protein of interest. On the other hand, if the goal is to quantitate
A
the actual degree of fibrosis, then normalizing a fibrotic marker to a constant amount of sample protein could be appropriate. Obtaining relative abundance data usually requires accumulating a sufficient number of samples per group and, ideally, analyzing the groups on a single blot (with multiple gels on one blot paper if necessary) so that all the samples to be compared are incubated with the same aliquots of primary and secondary antibodies, and quantified simultaneously. These considerations reduce the variables that interfere with quantifying relative abundance in the groups. If all the samples cannot be collected and prepared on the same day, then the sample handling between collection and assay becomes an important variable that warrants serious consideration. Often tissues are quick frozen and then prepared for assay together. Figure 1B illustrates the results of a comparison of samples of mouse kidney homogenates prepared from kidneys quick frozen in liquid nitrogen and stored for a month at ⫺80°C, versus kidney homogenates prepared immediately after euthanizing the mice. Results are shown for the renal Na⫹/H⫹ exchanger isoform 3 (NHE3, the sodium transporter responsible for the bulk of the renal sodium reabsorption). The apparent abundance of NHE3 is less in the frozen versus the freshly prepared homogenates, indicating that there has been a partial loss of the epitope (perhaps by degradation or masking) as a consequence of freezing and thawing. The mouse samples, in our opinion, are still useful as long as they are prepared identically (all freshly prepared or all frozen for about the same time). However, our experience indicates that frozen rat kidneys do not fare as well as frozen mouse kidneys. Figure 1C illustrates samples from frozen and fresh rat kidneys prepared with the same reagents
B
C frozen
fresh
Normal myocardium
mw (kD)
mw (kD)
frozen
fresh
100
NHE3
100
NHE3
70 70
250
NKCC
55 100 70
Infarcted myocardium
250
NCC mouse kidney homogenate 45 µg/lane
100 70
rat kidney homogenate 40 µg/lane Fig. 1. Consider sample integrity. A: illustration showing that the percentage of normal cardiac myocytes can be significantly decreased in infarcted myocardium versus normal myocardium. B: illustration of the effect of freezing (versus not freezing) mouse kidneys before homogenization on the detection of the Na⫹/H⫹ exchanger isoform 3 (NHE3) epitope, a measure of the apparent abundance of NHE3. C: illustration of the near total loss of renal transporter epitope recognition when rat kidneys are frozen before preparation of homogenates. See text for details and discussion. NKCC, Na⫹-K⫹-2 Cl⫺ cotransporter; NCC, Na⫹-Cl⫺ cotransporter. AJP-Cell Physiol • doi:10.1152/ajpcell.00400.2014 • www.ajpcell.org
Review C428
QUANTITATING IMMUNOBLOTS
on the same day and analyzed on the same blot. While there are strong signals for the NHE3, the Na⫹-K⫹-2 Cl⫺ cotransporter (NKCC), and Na⫹-Cl⫺ cotransporter (NCC) in the freshly prepared homogenates, the signals are lost in kidneys that were freeze/thawed before homogenate preparation. For this reason, we recommend that the investigator always compare homogenates made from freshly isolated tissue to frozen tissue to understand if there is serious loss of antibody epitope for their protein of interest. Interestingly, we have found that once renal homogenates are prepared with protease and phosphatase inhibitors (19) and stored in single-use aliquots at ⫺80°C, the immunoblot signals of renal transporters do not appear to further decay over many years of storage, whether the samples were originally from fresh or freeze/thawed tissue (not shown). Many sample handling factors can affect relative abundance. As two examples, the phosphorylation status of the renal NCC is profoundly influenced by the time since consuming the last meal (25), and we have now been made aware of the influence of circadian rhythms on protein expression (23). In addition, sample enrichment by subcellular fractionation or affinity purification will exhibit batch-to-batch variability in relative recoveries; analysis of homogenates circumvents this recovery consideration. In summary, the investigator needs to consider the status of the tissues being compared and assess how handling between collection and assay can influence the calculation of a sample or treatment groups’ relative abundance.
Consider the Specificity of the Antibody for the Target It is the investigator’s responsibility to provide details about each antiserum sufficient for a reader to replicate the immunodetection. This requires providing not just the vendor information but also catalog number, because the vendor can have multiple antibodies to one protein. Information about antiserum dilution, host, and specific secondaries are also very useful because antibody vendors come and go while an antisera can be used by an author over decades. If many antisera are used, these can be organized in a table. Does the antiserum recognize the protein of interest? It is the investigator’s responsibility to answer this question. It goes without saying that every immunoblot needs to include molecular weight standards to assess and report the apparent molecular weight of the protein(s) detected by the antiserum. It is important to acknowledge that just because a vendor advertises that an antiserum recognizes a target protein does not guarantee that the antibody detects that protein in your samples, nor that it does not also bind to another unrelated protein (nonspecifically) with a similar molecular weight. As a case in point, a recent study by Herrera, and a companion commentary by the Eguchi lab, provide proof that seven commercially available antibodies to the angiotensin type 1 receptors (AT1R) lack specificity (7, 11). Bands of the “expected” apparent molecular weight were equally evident in tissues and cells that expressed no AT1R as in tissues known to express AT1R. The investiN-terminal antibody detects full length SPAK
Fig. 2. Consider the specificity of the antibody for the target. SPAK [sterile 20 (STE20)/SPS1-related proline/alanine-rich kinase] was detected first with an anti-NH2-terminal epitope antiserum, which detects full-length SPAK (FL-SPAK), then a red-tagged secondary; the blot was reprobed with an anti-COOH-terminal epitope antisera that detects FL-SPAK, smaller SPAK 2, and an even smaller kidney-specific SPAK isoforms (KS-SPAK), then a green-tagged secondary. Bottom: illustration of the image obtained when both signals were detected at the same time; yellow indicates detection at the same location. SPAK is expressed in testes as a single full-length isoform (FL-SPAK) included as a positive control, is absent in a SPAK knockout mouse (SPAK-KO), included as a negative control, and all three isoforms are present in wild-type mouse kidney. Nonspecific background bands (white arrows) were identified based on expression in the SPAK-KO. See text for details and discussion.
C-terminal antibody detects SPAK isoforms
SPAK double labeling FL-SPAK (yellow)
FL-SPAK (yellow) background
SPAK 2 (green) KS-SPAK (green)
background
KS-SPAK (green) testis
SPAK-KO
AJP-Cell Physiol • doi:10.1152/ajpcell.00400.2014 • www.ajpcell.org
WT
WT
Review C429
QUANTITATING IMMUNOBLOTS
phorylated form, and perhaps other bands that are in compartments that do not become phosphorylated, as we observed for renal NCC using nondenaturing gels (13). When using antiphospho-specific antisera, the main concern is that the antiserum may also cross react with the nonphosphorylated protein. To avoid this, it is recommended to preabsorb the antiserum with a nonphosphorylated version of the immunizing peptide and to keep the primary antiserum incubation time short to minimize cross-reaction with the nonphosphorylated protein. In summary, the investigator needs to convincingly establish the specificity of the antiserum for their protein of interest by assessing whether it is detected at the correct apparent molecular size (by providing a stained lane of molecular weight markers on the immunoblot); by determining whether it is present in a known positive control (often available for purchase from vendors); by determining whether it is absent in a sample that does not contain the protein of interest; and by identifying nonspecific “background” bands. Consider the Linearity of the Detection System How much protein should be assayed? Assuming that the integrity of the sample and specificity of the antiserum for the
A
2.5
α1
relative abundance (60 µg = 1)
gators postulate that this may be a common problem with G protein-coupled receptors and provide the opinion that authors (and vendors) must apply rigorous standards to establish the specificity of the antibody for the target. Given the significant effort required to characterize an antibody, as well as potential pitfalls, unbiased resources have begun to be established on the web to aid investigators in choosing antibodies that have a good chance of working based on previous published reports (e.g., www.Citeab.com) and epitope sequence considerations (e.g., www.uniprot.org). We provide a typical characterization for the transporteractivating kinase known as SPAK [sterile 20 (STE20)/SPS1related proline/alanine-rich kinase], using antibodies to COOH- and NH2-terminal epitopes, in Fig. 2. These antibodies, prepared by the Delpire lab for use in studying brain SPAK (20, 21), were applied by us to the study of kidney SPAK. Unlike brain SPAK, in the kidney, SPAK is expressed as three nested isoforms sharing a common COOH terminus: fulllength SPAK (FL-SPAK), SPAK 2, and a smaller kidneyspecific form (KS-SPAK) (15). The Delpire lab generously provided our lab with both antibodies as well as a kidney from a SPAK knockout (KO) mouse, to serve as a negative control. Initial characterization of the antibodies (20) revealed very strong SPAK signals in testis, which serves as a positive control. The top panel (Fig. 2) displays the immunoblot results obtained after probing with the anti-NH2-terminal epitope antibody, present only on FL-SPAK, followed by a secondary antibody with a red label: The testis sample lane gives a strong positive FL-SPAK signal, there is no staining in the SPAK KO kidney lane in the FL-SPAK region, and strong bands are evident at the position of FL-SPAK in the two wild-type (WT) mouse kidney sample lanes. However, there are three prominent bands below the FL-SPAK band coincident in both the WT and KO samples, indicating that these are nonspecific “background” bands. The middle panel displays the results obtained when the same blot was probed with an anti-COOHterminal epitope antiserum expected to detect all three of the nested renal SPAK isoforms, followed by a secondary antibody with a green label: A single strong band is evident in the positive control lane containing testis (which expresses only FL-SPAK), only a very faint background band is evident in the SPAK KO lane, indicating that this antisera is quite specific, and multiple bands are evident in the two WT lanes. The two images were superimposed, which indicates that 1) the FLSPAK, recognized by both the anti-COOH- and anti-NH2terminal antisera, appears yellow (indicating that they both recognize the same protein) in testis and the WT samples; 2) the SPAK 2 and KS-SPAK, which are recognized by only the COOH-terminal probe, appear green; and 3) “background bands,” detected by the NH2-terminal antibody, appear red. This example illustrates the utility of using fluorescent tagged secondary antibodies to multiple epitopes to discriminate between specific and non-specific labeling. With this information we were able to identify SPAK, SPAK 2 and KS-SPAK and determine how their levels were regulated by angiotensin II as well as nitric oxide synthase inhibition (8, 9). Fluorescent tagged secondaries are also useful to analyze protein phosphorylation because, if they are generated in two different species, one color can be used to detect the total protein and another for the phosphorylated form, and the overlay should detect a band labeled with both for the phos-
α2
2.0
α3 NCE
1.5
20
40
60
(µg)
80
100
1.0
0.5
0.0 0
20
40
60
80
100
protein loaded/lane (µg)
B
rat renal cortex
rat renal medulla
NKA α1 0.12
0.25
0.5
24
48
0.12 0.25
0.5 µg/lane
NKA β1 3
6
12
3
6
12
24
48 µg/lane
Fig. 3. Consider the linearity of the detection system. A: illustration of the immunodetection of sodium pump catalytic isoform subunits (␣1, ␣2, ␣3) and the Na⫹/Ca2⫹ exchanger (NCE) from human heart detected between 20 and 100 g/lane. [Redrawn from Wang et al. (27).] B: illustration of the immunodetection of sodium pump (NKA) ␣1-subunit between 0.12 and 0.5 g/lane and 1-subunit between 3 and 48 g/lane in rat kidney homogenates from cortex and medulla. See text for details and discussion.
AJP-Cell Physiol • doi:10.1152/ajpcell.00400.2014 • www.ajpcell.org
Review C430
QUANTITATING IMMUNOBLOTS
protein of interest are satisfactorily characterized, the investigator and critical reader should consider whether the conditions of sample loading and quantitation provide signals in the linear range of the detection system (regardless of whether the system is detecting light or radioactivity or fluorescence). That is, one should determine whether the signal generated from the primary and secondary antibody labeling is a direct measure of the relative abundance of the protein of interest in the samples. It is our opinion that linearity of the detection system should be established for each protein of interest in every study (ideally in every immunoblot) because of the high number of variables that can influence the final signal. Variables include the level of protein expression in the tissue, sample handling and preparation (Fig. 1), gel transfer conditions, blocking efficiency, antibody dilution, affinity for the target, and detection method (e.g., chemiluminescence emits a signal that changes with time); we contend that significant regulation is often missed or inaccurately assessed because immunoblots are analyzed without assessing linearity. We propose that investigators analyze multiple amounts of samples on the same blot to calibrate detection; Charette et al. (6) propose a similar strategy emanating from their analysis of nonlinear chemiluminescent signals. The inset of Fig. 3A provides a series of immunoblots of cardiac homogenates in which samples were loaded at 20, 40, 60, 80, and 100 g/lane then individually probed with antisera directed against sodium pump catalytic subunit isoforms (␣1, ␣2, and ␣3) or against the sodium/calcium exchanger (NCE) (redrawn from Ref. 27). The graph (Fig. 3A) plots relative abundance of the immunoblot signal from each protein of interest, normalized to the abundance of the signal from the 60 g sample defined as 1.0; the dotted line indicates equivalent
changes in protein loaded and density signal. Between 20 and 80 g/lane, the change in relative signal abundance is roughly equivalent to the amount of protein loaded for ␣2, ␣3, and NCE, but not for ␣1. This is a good illustration of a problematic issue because while the density of the ␣1 signal is “light,” compared with the NCE signal, the detection system is saturated above 60 g for ␣1, but not for NCE. Said another way, just because signals on an immunoblot are in the gray zone does not guarantee they are in a linear range of detection—a doubling of signal with abundance must be validated for the specific assay conditions. Figure 3B compares detection of renal sodium pump subunits (␣1 and 1). Despite the fact that the analysis shown is restricted to a linear range, it is clear that this linear range of detection for kidney ␣1 is attained with about 100-fold less homogenate than found for cardiac tissue, corresponding to the far greater abundance of Na-K-ATPase ␣1 in kidney versus heart. While this finding is not surprising, Fig. 3B also indicates that detection of the Na-K-ATPase 1 subunit in renal tissue is linear when loaded at 100-fold more than that for detection of ␣1. Why are the ranges not more similar for this ␣ heterodimer? One explanation is that 1 is a glycoprotein and the sugar residues reduce antibody-antigen binding; we and others established long ago that  detection improves when  is deglycosylated before immunodetection (3). Another explanation that needs to be considered for all immunodetection is that an antisera’s affinity for an antigen-containing protein is dependent on the biological reaction of the immunized host against the injected antigen during antisera production and, thus, is highly variable between hosts or hybridoma clones.
A 98.9
92.6
92.7
103.0
85.1
45.3
47.6
56.9
64.6
67.8
28.7
26.3
27.6
27.4
29.8
23.9
26.1
27.3
28.7
26.0
6µg/lane:
Fig. 4. Consider the loading controls: part 1. A: illustration of a single immunoblot constructed from two SDS-PAGE gels loaded with 6 g/lane (top of blot) and 3 g/lane (bottom part of blot) kidney homogenate and probed with antiserum to claudin 10 and a secondary antiserum tagged red. The quantitation, indicated to the right of the blot, demonstrates linearity of the detection system: Specifically, signal intensity increases 1.7-fold with sample doubling. B: image of the same blot reprobed with an anti-actin (pan-actin) antiserum and a secondary antiserum tagged green. The actin signal intensity does not change with sample doubling. We conclude that actin is not an acceptable loading control for this application; rather it is “overloaded” at 3 and 6 g kidney homogenate/lane. See text for further details and discussion.
3µg/lane:
B 6µg/lane:
3µg/lane:
AJP-Cell Physiol • doi:10.1152/ajpcell.00400.2014 • www.ajpcell.org
Review C431
QUANTITATING IMMUNOBLOTS
Consider the Loading Controls Is a loading control necessary if one has established the linearity of the detection system? Yes. Equivalent loading of sample amount per lane is important to establish if immunoblots will be quantitated. Unequal sample-to-sample loading can result from: high levels of variability in the determination of protein concentration of the samples, incorrect sample preparation, or poor pipetting into the wells. It is very common for investigators to reprobe their blots with what is known as a loading control or housekeeping protein (e.g., actin or GAPDH) that exhibits high-level, constitutive expression in the tissue and is not expected to vary with the treatment. However, a key problem with this approach is that these loading control proteins are rarely in the same linear range of detection as the target protein of interest. A number of studies concur that in the typical loading range of 10 –50 g of cell lysate, quantitation of housekeeping proteins by immunodetection is not possible because they are in saturating quantities (1, 10, 28). A good illustration of the issue is provided in Fig. 4 in which five samples of renal homogenates were resolved at 6 and 3 g/lane (to establish linearity). The immunoblot was first probed with an antiserum directed to claudin 10 (a tight junction protein) and a red-tagged secondary (Fig. 4A, left) and claudin 10 was detected between the 17 and 28 kDa molecular mass markers. Quantitation of the five samples, provided to the right, indicates a good linearity of the detection system as the signals were roughly doubled with 6 versus 3 g protein loading. The same blot was probed with an antiserum to actin and a green-tagged secondary, and actin was detected above 34 kDa (Fig. 4B). The signal intensity of the actin band does not significantly change when sample loaded was reduced to ½; thus, actin is not an acceptable loading control for this application; it is saturating the membrane at just 3 g/lane. An alternative to a single loading control is protein staining of the gel or blot as a loading control (1, 28). Vendors are starting to market kits to stain loading gels, but it can be done
with what is on hand in the lab. We make up enough of each sample to run a parallel “loading gel” which is loaded and run identically to the gel(s) that will be immunoblotted (another approach is to actually stain the blot but we are concerned that this could interfere with immunodetection). In our approach, used by others as well, the gel is not blotted, rather, it is simply stained and destained and a number of bands are chosen to quantitate by densitometry. Another approach is to quantitate the density of the entire lane from top to bottom, but we have had good results assessing discrete bands that do not appear to be too dark or too light. Figure 5 provides an illustration of a loading gel containing eight samples that were all, nominally, loaded with 7.5 g/lane of renal cortex homogenate. Five bands were identified, quantitated by densitometry, then normalized to the mean density for that band in all eight samples, defined as 1.0. The relative abundance results for each sample as well as the average value for the five bands within each sample are displayed below the loading gel (Fig. 5). A visual inspection of the loading gel suggests samples 4 and 7 are relatively overloaded and the quantitation bears this out, as both have relative average densities of 1.13; the others vary by 5% or less. At this point, the investigator can report the level of variability that is acceptable for the quantitative immuno-
1
2 3
4 5
Loading gel- 7.5 µg renal cortex/lane 1.2
Relative Density
In summary, if an investigator aims to quantify changes in protein expression (or phosphorylation) using immunoblots, it is our experience and opinion that each immunoblot should contain a demonstration of the linearity of the detection system. This can be accomplished in a number of ways. We routinely run samples at 1 and ½ protein amounts (which may require multiple gels that are transferred together on the same blot membrane) and then process the single blot through immunodetection using the identical primary and secondary antibodies and simultaneous quantitation (as in Fig. 4). We calculate the linearity ratio by dividing the signal obtained with loading of the full amount with signal obtained with ½ the amount and expect it to be close to 2. If not, then the assay is underestimating or overestimating the effect of the treatment. In some cases we have included blots containing both 1 and ½ loading amounts in our figures (19) and in others we have reported the method used to establish linearity and shown just one amount in the figures (9). In this latter case, we have been asked by reviewers to provide images of every immunoblot loaded at the two amounts for verification. Another option is to provide a linearity analysis (as in Fig. 3A) as a supplemental figure, if the journal allows.
Band 1 1.1
Band 2 Band 3
1.0
Band 4 Band 5
0.9
AVERAGE 0.8 0
1
2
4
3
5
6
7
8
sample #
Fig. 5. Consider the loading controls: part 2. Direct protein staining of the gel or blot is an alternative to the loading control to assess whether the same amount of protein is loaded per lane. Top: image of a stained post SDS-PAGE gel (not blot) of renal homogenate samples from eight rat kidneys loaded at 7.5 g/lane. Arrows indicate the 5 unidentified protein bands arbitrarily chosen for quantitation to determine whether the 8 samples were equivalently loaded. Values were normalized to the mean value of each band in the 8 samples defined as 1.0. Bottom: illustration of the relative density of each of the 5 arbitrary bands (color coded) in each sample as well as the average of the 5 bands (black). The results indicate that samples 1, 2, 3, 6, and 8 are very close to equivalent with average relative density of 1.0 while 4, 5, and 7 deviate, by ⬍13%. See text for further details and discussion.
AJP-Cell Physiol • doi:10.1152/ajpcell.00400.2014 • www.ajpcell.org
Review C432
QUANTITATING IMMUNOBLOTS
blots, and options for proceeding include correcting the density of the immunoblots for this variance, beginning again with the protein assays and sample preparation to reduce sample to sample variability to 5–10% from the mean average density or less, or reporting and accepting the variability after providing a rationale. In summary, staining the gel provides a direct image of the actual protein loaded. The stained gel approach is a far more valuable loading control than that obtained by quantitating saturated signals from housekeeping proteins. The latter reveals little more than whether the sample was loaded or not. Conclusion During the past 35 years, immunoblotting has been a critical tool for making discoveries and progress in many areas of investigation from early use in expression cloning of genes, to utilization for studies of protein expression, to screening antibodies and clinical immunodiagnostics. Physiologists, with their focus on how tissues and organ systems work normally and in response to challenges, have relied on the Western blot’s ability to detect a “needle in a haystack” to address questions about organ system operation and regulation. However, since the method is simple to set up and implement, results are often conducted and reported without the requisite controls. Samples and reagents must be validated and controlled to obtain interpretable results. Said another way, if an investigator aims to determine whether there are more needles in one haystack than another, it is key to convincingly validate the integrity of the sample, the specificity of the antibody, and the linearity of the detection system and to assess sample loading. NOTE ADDED IN PROOF We recently became aware of a very useful review by Murphy and Lamb (18a) that similarly reviews important considerations for quantitating proteins in muscle using antibody-based techniques. ACKNOWLEDGMENTS Drs. Jo Adams and Christina Bennett provided very helpful comments. GRANTS This work was supported by National Institute of Diabetes and Digestive and Kidney Diseases Grant DK-083785 (to A. A. McDonough). DISCLOSURES No conflicts of interest, financial or otherwise, are declared by the author(s). AUTHOR CONTRIBUTIONS A.A.M., L.C.V., J.N.M., and D.L.R. performed experiments; A.A.M. prepared figures; A.A.M. drafted manuscript; A.A.M., L.C.V., J.N.M., and D.L.R. edited and revised manuscript; A.A.M., L.C.V., J.N.M., and D.L.R. approved final version of manuscript. REFERENCES 1. Aldridge GM, Podrebarac DM, Greenough WT, Weiler IJ. The use of total protein stains as loading controls: an alternative to high-abundance single-protein controls in semi-quantitative immunoblotting. J Neurosci Methods 172: 250 –254, 2008. 2. Alegria-Schaffer A, Lodge A, Vattem K. Performing and optimizing Western blots with an emphasis on chemiluminescent detection. Methods Enzymol 463: 573–599, 2009. 3. Azuma KK, Hensley CB, Tang MJ, McDonough AA. Thyroid hormone specifically regulates skeletal muscle Na⫹-K⫹-ATPase ␣2- and 2-isoforms. Am J Physiol Cell Physiol 265: C680 –C687, 1993.
4. Burnette WN. Western blotting: remembrance of past things. Methods Mol Biol 536: 5–8, 2009. 5. Burnette WN. “Western blotting”: electrophoretic transfer of proteins from sodium dodecyl sulfate–polyacrylamide gels to unmodified nitrocellulose and radiographic detection with antibody and radioiodinated protein A. Anal Biochem 112: 195–203, 1981. 6. Charette SJ, Lambert H, Nadeau PJ, Landry J. Protein quantification by chemiluminescent Western blotting: elimination of the antibody factor by dilution series and calibration curve. J Immunol Methods 353: 148 –150, 2010. 7. Elliott KJ, Kimura K, Eguchi S. Lack of specificity of commercial antibodies leads to misidentification of angiotensin type-1 receptor protein. Hypertension 61: e31, 2013. 8. Giani JF, Janjulia T, Kamat N, Seth DM, Blackwell WL, Shah KH, Shen XZ, Fuchs S, Delpire E, Toblli JE, Bernstein KE, McDonough AA, Gonzalez-Villalobos RA. Renal angiotensin-converting enzyme is essential for the hypertension induced by nitric oxide synthesis inhibition. J Am Soc Nephrol 25: 2752–2763, 2014. 9. Gonzalez-Villalobos RA, Janjoulia T, Fletcher NK, Giani JF, Nguyen MT, Riquier-Brison AD, Seth DM, Fuchs S, Eladari D, Picard N, Bachmann S, Delpire E, Peti-Peterdi J, Navar LG, Bernstein KE, McDonough AA. The absence of intrarenal ACE protects against hypertension. J Clin Invest 123: 2011–2023, 2013. 10. Hammond M, Kohn J, Oh K, Piatti P, Liu N. A method for greater reliability in Western blot loading controls: stain-free total protein quantitation. Bio-Rad Bulletin 6360, 2013. 11. Herrera M, Sparks MA, Alfonso-Pecchio AR, Harrison-Bernard LM, Coffman TM. Lack of specificity of commercial antibodies leads to misidentification of angiotensin type 1 receptor protein. Hypertension 61: 253–258, 2013. 12. Kurien BT, Scofield Hal R. Protein Blotting and Detection. Methods Mol Biol 536: 2009. 13. Lee DH, Maunsbach AB, Riquier-Brison AD, Nguyen MT, Fenton RA, Bachmann S, Yu AS, McDonough AA. Effects of ACE inhibition and ANG II stimulation on renal Na-Cl cotransporter distribution, phosphorylation, and membrane complex properties. Am J Physiol Cell Physiol 304: C147–C163, 2013. 14. Mahmood T, Yang PC. Western blot: technique, theory, and trouble shooting. N Am J Med Sci 4: 429 –434, 2012. 15. McCormick JA, Mutig K, Nelson JH, Saritas T, Hoorn EJ, Yang CL, Rogers S, Curry J, Delpire E, Bachmann S, Ellison DH. A SPAK isoform switch modulates renal salt transport and blood pressure. Cell Metab 14: 352–364, 2011. 16. McDonough A. Immunodetection of Na,K-ATPase in guinea-pig retinal layers, cornea and lens. Exp Eye Res 40: 667–674, 1985. 17. McDonough A, Schmitt C. Comparison of subunits of cardiac, brain, and kidney Na⫹-K⫹-ATPase. Am J Physiol Cell Physiol 248: C247– C251, 1985. 18. McDonough AA, Hiatt A, Edelman IS. Characteristics of antibodies to guinea pig (Na⫹ ⫹ K⫹)-adenosine triphosphatase and their use in cell-free synthesis studies. J Membr Biol 69: 13–22, 1982. 18a.Murphy RM, Lamb GD. Important considerations for protein analyses using antibody based techniques: down-sizing Western blotting up-sizes outcomes. J Physiol 591: 5823–5831, 2013. 19. Nguyen MT, Yang LE, Fletcher NK, Lee DH, Kocinsky H, Bachmann S, Delpire E, McDonough AA. Effects of K⫹-deficient diets with and without NaCl supplementation on Na⫹, K⫹, and H2O transporters’ abundance along the nephron. Am J Physiol Renal Physiol 303: F92–F104, 2012. 20. Piechotta K, Garbarini N, England R, Delpire E. Characterization of the interaction of the stress kinase SPAK with the Na⫹-K⫹-2Cl⫺ cotransporter in the nervous system: evidence for a scaffolding role of the kinase. J Biol Chem 278: 52848 –52856, 2003. 21. Piechotta K, Lu J, Delpire E. Cation chloride cotransporters interact with the stress-related kinases Ste20-related proline-alanine-rich kinase (SPAK) and oxidative stress response 1 (OSR1). J Biol Chem 277: 50812–50819, 2002. 22. Renart J, Reiser J, Stark GR. Transfer of proteins from gels to diazobenzyloxymethyl-paper and detection with antisera: a method for studying antibody specificity and antigen structure. Proc Natl Acad Sci USA 76: 3116 –3120, 1979.
AJP-Cell Physiol • doi:10.1152/ajpcell.00400.2014 • www.ajpcell.org
Review QUANTITATING IMMUNOBLOTS 23. Richards J, Gumz ML. Mechanism of the circadian clock in physiology. Am J Physiol Regul Integr Comp Physiol 304: R1053–R1064, 2013. 24. Schmitt CA, McDonough AA. Developmental and thyroid hormone regulation of two molecular forms of Na⫹-K⫹-ATPase in brain. J Biol Chem 261: 10439 –10444, 1986. 25. Sorensen MV, Grossmann S, Roesinger M, Gresko N, Todkar AP, Barmettler G, Ziegler U, Odermatt A, Loffing-Cueni D, Loffing J. Rapid dephosphorylation of the renal sodium chloride cotransporter in response to oral potassium intake in mice. Kidney Int 83: 811–824, 2013.
C433
26. Towbin H, Staehelin T, Gordon J. Electrophoretic transfer of proteins from polyacrylamide gels to nitrocellulose sheets: procedure and some applications. Proc Natl Acad Sci USA 76: 4350 –4354, 1979. 27. Wang J, Schwinger RH, Frank K, Muller-Ehmsen J, Martin-Vasallo P, Pressley TA, Xiang A, Erdmann E, McDonough AA. Regional expression of sodium pump subunits isoforms and Na⫹-Ca⫹⫹ exchanger in the human heart. J Clin Invest 98: 1650 –1658, 1996. 28. Welinder C, Ekblad L. Coomassie staining as loading control in Western blot analysis. J Proteome Res 10: 1416 –1419, 2011.
AJP-Cell Physiol • doi:10.1152/ajpcell.00400.2014 • www.ajpcell.org
