The use of knockout mice to test the specificity of antibodies for cannabinoid receptors

Suarez et al. (2009) characterized immunolabeling in the brain for both cannabinoid CB1 and CB2 receptors. Various researchers have observed that obtaining robust results for immunohistochemistry has been challenging for G-protein couple receptors. Reliable antibodies for cannabinoid receptors have proven especially difficult to construct (Grimsey et al., 2008) and immunohistochemistry for the cannabinoid CB2 receptor has been the focus of recent controversy (Atwood and Mackie, 2010). Suarez et al. (2009) presented evidence for the specificity of the antibodies used in their study, using the method of testing the antibodies on sections taken from transgenic mice lacking either cannabinoid receptor. So called ‘‘knockout (KO) controls’’ are considered the best available test for antibody specificity (Lorincz and Zoltan, 2008). However, the evidence for the specificity of the CB2 antibody used by Suarez et al. (2009) may be questioned. When the images from Figure 1 in Suarez et al. (2009) were imported into Adobe Photoshop V5.0 (Abobe, USA) and the input levels of the knockout controls adjusted to match the mean and range of the output levels of the histogram for corresponding WT section, striking differences between the CB1 and CB2 KO controls emerged. This adjustment equalizes the mean and range of intensity of the KO image in the black channel with the WT section, so that the images only differ with respect to the pattern but not intensity of staining. This method determines whether immunolabeling patterns have been changed in the KO section (as expected by removal of the specific binding site) or simply reduced in intensity. This revealed that the CB2 antibody had a near identical pattern of staining in KO sections as in wild-type (WT) sections, but at reduced intensity. By contrast, the CB1 antibody gave a qualitatively different pattern of staining in the KO section compared to WT. Tellingly, this CB1 antibody has been extremely well validated is other assays, and is the only one of a large number of antibodies tested in a recent study of CB1 antibody specificity that yields a pattern of CB1 expression consistent with autoradiography (Grimsey et al., 2008). When level histograms for the equalized images are compared, it can be seen that the frequency distribution for the CB1 KO is of a different shape than for WT, as expected if specific binding sites are removed by the gene deletion (Fig. 1). However, the frequency distribution for the CB2 KO is very similar in shape to the distribution for the WT image. Therefore, both visual inspection and spectral analysis show that the CB2 KO image has the same pattern of staining as WT, but at reduced intensity. This result is incompatible with a high degree of specificity for the CB2 antibody. Immunolabeling in a knockout control section should demonstrate an absence (not merely a reduction in number or ‘‘knockdown’’) of the specific binding site. Therefore, background staining in KO sections should show a qualitatively different pattern than that found in WT sections. We propose three possible explanations for the results for the CB2 receptor in the KO test in Suarez at al. (2009). First, partial expression of the CB2 receptor may occur in the CB2 KO mouse. This might be explained by the recent discovery of isoforms of the CB2 receptor (Liu et al., 2009). However, the relevant C-terminus region of the mouse receptor is identical in the two known isoforms, and differences in the mice CB2 isoforms only exist in promoter regions of the gene, not in the protein coding regions. Therefore, the epitope for this Cterminus antibody should be absent from the C-terminus KO mouse for both known CB2 isoforms. Second, it is conceivable that the specific binding site for the CB2 antibody precisely overlaps with low-affinity nonspecific binding sites and with background staining, such that addition of specific binding merely intensifies the staining pattern for nonspecific staining. Arguing against this explanation is the qualitatively different pattern of nonspecific and background staining seen in the CB1 KO mouse for the CB1 antibody. A third possibility is that the CB2 KO and WT sections have been subject to slight differences in the DAB staining protocol due to small random variations in such parameters as ambient light, temperature, incubation time, and other factors that influence DAB staining. Peroxidase-based DAB staining is highly sensitive to small differences between protocols, and it is crucially important that multiple negative control sections are run at the same time as experimental sections. Irrespective of which of these explanations is the correct, the specificity of the CB2 receptor antibody used by Suarez et al. (2009) is open to doubt. Department of Pharmacology & Toxicology, Otago School of Medical Sciences, University of Otago, PO Box 913, Dunedin, New Zealand Correspondence to: John C. Ashton, Department of Pharmacology & Toxicology, Otago School of Medical Sciences, University of Otago, PO Box 913, Dunedin 9054, New Zealand. E-mail: john.ashton@otago.ac.nz Accepted for publication 17 February 2011 DOI 10.1002/hipo.20946 Published online 2 May 2011 in Wiley Online Library (wileyonlinelibrary.com). HIPPOCAMPUS 22:643–644 (2012)