Specific detection of interferon regulatory factor 5 (IRF5): A case of antibody inequality

Interferon regulatory factor 5 (IRF5) is a member of the IRF family of transcription factors. IRF5 was first identified and characterized as a transcriptional regulator of type I interferon expression after virus infection. In addition to its critical role(s) in the regulation and development of host immunity, subsequent studies revealed important roles for IRF5 in autoimmunity, cancer, obesity, pain, cardiovascular disease, and metabolism. Based on these important disease-related findings, a large number of commercial antibodies have become available to study the expression and function of IRF5. Here we validate a number of these antibodies for the detection of IRF5 by immunoblot, flow cytometry, and immunofluorescence or immunohistochemistry using well-established positive and negative controls. Somewhat surprising, the majority of commercial antibodies tested were unable to specifically recognize human or mouse IRF5. We present data on antibodies that do specifically recognize human or mouse IRF5 in a particular application. These findings reiterate the importance of proper controls and molecular weight standards for the analysis of protein expression. Given that dysregulated IRF5 expression has been implicated in the pathogenesis of numerous diseases, including autoimmune and cancer, results indicate that caution should be used in the evaluation and interpretation of IRF5 expression analysis.

The interferon regulatory factor 5 (IRF5) gene encodes proteins that bind to DNA and regulate the transcription of genes important for innate and adaptive immunity, as well as cell growth regulation and apoptosis [1][2][3][4][5][6][7] . IRF5 was originally cloned from an expressed sequence tag (EST) human dendritic cell cDNA library and human B cells 1 . The first analysis of its expression was by Northern blot using a commercially available human healthy donor tissues blot. Constitutive IRF5 mRNA expression was detected primarily in lymphoid tissues and peripheral blood lymphocytes (PBL) with the highest levels detected in spleen and PBL 1 . In all samples identified to express IRF5, two transcripts were detected. Subsequent findings from multiple labs have now shown that IRF5 exists as multiple alternatively spliced transcripts that encode distinct isoforms 3 . IRF5 expression was also analyzed in human tumor cell lines and those cell lines expressing IRF5 showed only one transcript 1,3 . Notably, all immortalized lymphoma and leukemia cell lines examined at that time showed the absence of IRF5 mRNA expression.
To date, IRF5 is best known for its 1) central role in pathogen-induced immunity via activation by the MyD88-dependent Toll-like receptor (TLR) signaling pathway 8,9 , 2) identification as an autoimmune susceptibility gene [10][11][12][13][14][15][16][17][18][19] , and 3) tumor suppressor function 4,6,7,20,21 . With the availability of Irf5 transgenic knockout mice, however, IRF5 has been also shown to play important roles in macrophage polarization, pain management, obesity, cardiovascular disease, systemic lupus erythematosus (SLE), arthritis and metabolic dysfunction 19,[22][23][24][25][26][27][28][29][30][31] . IRF5 proteins generally reside in the cytoplasm of a quiescent cell and upon stimulation (with virus, TLR ligands or DNA damage, etc.) become activated via post-translational modification(s), resulting in nuclear translocation and binding to the promoters of target genes 2 . IRF5's function (or dysfunction) in autoimmune disease and cancer is tightly linked to its expression. In the case of autoimmune diseases such as SLE, polymorphisms in the IRF5 gene, referred to as risk polymorphisms, result in, or associate with, elevated IRF5 expression 11,[32][33][34] . Conversely, in the case of cancer, IRF5 expression is often found to be downregulated (or absent) in malignant versus non-malignant cells. More recent data on the analysis of IRF5 expression and function in different cancer types has revealed that in some cases, such as in splenic marginal zone lymphoma, IRF5 expression is decreased 21 , while in Hodgkin's lymphoma, IRF5 expression is elevated 35 . The same may be true for autoimmune diseases as both risk and non-risk polymorphisms have been associated with susceptibility to and protection from, respectively, a particular disease. In either case, analysis of IRF5 expression by immune-based techniques, such as immnuoblot, flow cytometry, and immunohistochemistry, has become very important for our understanding of IRF5 function (or dysfunction) in a disease-specific context.
In the current study, we used established positive and negative controls for the analysis of IRF5 antibody specificity in human cells and tissues, and Irf5 gene knockout (ko) mice for analysis in murine cells and tissues. Data presented herein indicate that caution should be used in the interpretation of findings from particular antibodies as many are not specific for IRF5; a number of commonly used antibodies detect IRF5 in negative controls. We have identified a few key antibodies that are very good at the specific detection of IRF5 expression in particular applications. Given the clinical utility of IRF5 expression as a biomarker of disease susceptibility and severity, appropriate controls and molecular weight standards should be included and published in order to authenticate antibody specificity, as well as scientific findings.

Results
In primary leukocytes and lymphocytes, IRF5 expression is highest in monocytes and macrophages, lower in dendritic cells and B cells, and nearly undetectable in T cells and NK cells. For the validation of currently available commercial anti-IRF5 antibodies, three immortalized human lymphoid cell lines -THP1 monocytic macrophages, Ramos B cells, and Jurkat T cells -that express distinctly different levels of endogenous IRF5 transcript and protein expression were used. Expression differences were confirmed by real-time quantitative polymerase chain reaction (qRT-PCR) and immunoblot analysis using lysates from cell lines newly obtained from ATCC and commercially purchased lysates (Fig. 1A,B). Protein levels were confirmed with previously validated antibodies that are no longer available -Cell Signaling #3257 (cs3257) or clone 2E3-1A11 from Novus Biologicals (H00003663-M01) and Sigma (WH0003663M1). As expected, cs3257 detected high levels of IRF5 in protein lysates from THP1 cells 1,3 . Compared to THP1, Ramos B cells expressed dramatically lower levels, and Jurkat T cells, as most T cell lines, expressed little to no IRF5 (Fig. 1B). Surprisingly, we have found that some companies used Jurkat T cells as a positive control for IRF5 protein detection; to our knowledge, there have been no publications showing IRF5 expression in Jurkat cells. To the contrary, two recent publications have shown that Jurkat T cells lack IRF5 expression 36,37 . Furthermore, analysis of Jurkat RNAseq reads from ENCODE using the UCSC Genome Browser indicate that Jurkat cells either express too low IRF5 that reads are not detected across exons or expression is absent 38,39 .
Using protein lysates from the three cell lines, seven commercially available anti-IRF5 antibodies were tested for detection specificity by comparing to the previously available and validated cs3257 antibody. Key experimental factors, protein loading, antibody concentration, etc., were controlled for each antibody and are detailed in Fig. 1C. Although the majority of antibodies tested were able to detect a band (or bands) between the 72 and 55 kDa protein markers that correspond to the approximate size of endogenous IRF5 (~62 kDa), most antibodies were unable to distinguish IRF5 expression differences between cell lines (Fig. 1D). Antibodies #33478 (Abcam; ab33478) and #13496 (Cell Signaling; cs13496) detected similar, high IRF5 expression in all three cell lines indicating lack of specificity. Antibodies #124792 (Abcam; ab124792) and NB100-1092 (Novus Biologicals) detected a similar large, single band in THP1 cells but were unable to detect IRF5 in Ramos B cells; a faint band of appropriate size could be detected in lysates from Ramos B cells after longer exposures with ab124792 but not NB100-1092. While antibody #2932 (Abcam; ab2932) detected clean bands of appropriate size in all three cell lines, the levels of expression were incorrect showing Ramos B cells as the highest expressers of IRF5. Antibody SAB1403991 (Sigma) was the least specific as it detected multiple bands throughout the membrane and antibody #181533 (Abcam; ab181533) was the most specific since it cleanly detected expression differences between the cell lines. Ab181533 was the only antibody capable of detecting IRF5 expression differences between each cell line that correlated with transcript levels and detection with cs3257.
To further confirm antibody specificity, targeted siRNAs were used to knockdown endogenous IRF5 in Ramos B cells. qRT-PCR and immunoblot analysis with the validated cs3257 antibody were used to confirm IRF5 knockdown ( Fig. 1E and data not shown). Lysates from the same batch of nucleofected cells were used to generate multiple independent blots for comparison of antibody specificity. Similar to immunoblot data in panel D, antibodies ab33478 and ab2932 were unable to detect differences in IRF5 expression after knockdown, and ab124792 and NB100-1092 were unable to detect IRF5 in Ramos B cells at all. Interestingly, although multiple bands were detected with SAB1403991, one band of appropriate size disappeared after knockdown suggesting that the antibody may be able to recognize IRF5, albeit in a non-specific manner. Antibody cs13496 was also interesting as it detected IRF5 knockdown in Ramos B cells even though it was unable to distinguish expression differences between the three cell lines (Fig. 1D). However, the most specific antibody for IRF5 detection, before and after knockdown, was ab181533 (Fig. 1E). Ab181533 was the most consistent and specific in its ability to detect differing levels of endogenous IRF5 by immunoblot analysis.
We next evaluated the ability and efficiency of IRF5 antibodies to immunoprecipitate (IP) endogenous IRF5 proteins from Ramos B cell lysates. Even though antibodies NB100-1092 and ab2932 were not able to recognize IRF5 by immunoblot analysis, we tested them for IP as the companies specifically recommended them for this method on their technical sheets. Cs3257 and clone 2E3-1A11 (WH0003663M1 or H00003663-M01) were also used for IP comparison as antibodies from clone 2E3-1A111 were the only ones previously available that could IP endogenous IRF5. Protein lysates were immunoprecipitated with the indicated antibodies, as detailed in Fig. 2A, and IP efficiency assessed by immunoblot analysis with H00003663-M01. Interestingly, cs3257 was unable to IP much IRF5 while WH0003663M1 (clone 2E3-1A11) could. Given that WH0003663M1 and H00003663-M01 were both monoclonal, high background bands for heavy and light chain immunoglobulins were seen on the blot, in addition to a band corresponding to IRF5 (Fig. 2B). Not surprising, ab33478 and ab2932 were unable to IP IRF5, while ab124792 immunoprecipitated similar levels as the previously available WH0003663M1, yet with less non-specific binding. Since NB100-1092 was recommended by the manufacturer for IP, we performed an antibody titration to detect IP efficiency and specificity. Somewhat surprising given the immunoblot data in Fig. 1D,E, NB100-1092 was able to precipitate large amounts of IRF5 with 2 μ g of antibody; however, a number of other proteins precipitated with IRF5 indicating a lack of specificity (Fig. 2C). Ab181553 was also tested, given its sensitivity and selectivity for IRF5 by WB, and while it was able to IP IRF5, additional non-specific bands were detected (Fig. 2D). Since ab124792 gave the least non-specific binding (Fig. 2B), this antibody was further optimized by titration and washing after IP. We found highly efficient and specific IP of IRF5 at 4 μ g antibody (Fig. 2E). To further confirm specificity, ab124792 was used for IP in Jurkat T cells and no binding was detected at the same concentration of protein lysate and antibody used for IP of Ramos B cell lysates (Fig. 2F). Of interest, Abcam specifically states on their product sheet that ab124792 is unsuitable for IP. Based on the current analysis, we found ab124792 to be a specific and efficient antibody for IP of human IRF5.
Next, to determine whether any of the antibodies were capable of detecting intracellular IRF5 by flow cytometry analysis, as in Fig. 1, IRF5 expression was compared between THP1, Ramos, and Jurkat cells. Using the same antibody concentration, permeabilization protocol, and cell staining procedure (Fig. 3A), we found that antibodies cs3257, ab124792, cs13496, and SAB1403991 were able to detect the expected IRF5 expression differences between cell lines (Fig. 3B). Ab124792 was specifically recommended for flow cytometry on its technical sheet and it showed the largest difference between cell peaks (based on MFI values). While not used in the current study, ab124792 is now available in pre-conjugated forms (ab193245 or ab192983) that we have found to further enhance specificity for intracellular IRF5 detection by flow cytometry.
Given the importance of IRF5 as a potential biomarker for autoimmune disease and cancer, we evaluated the ability of antibodies to detect IRF5 by immunohistochemical (IHC) and immunofluorescence (IF) microscopy in human splenic tissue. Spleen was chosen for comparative analysis of antibody detection since previous data from Northern blot and qRT-PCR analysis of IRF5 transcript expression showed high levels of IRF5 in this organ 1 . Antibodies cs3257, ab124792 and ab181553 were chosen for testing on formalin-fixed paraffin-embedded (FFPE) spleen tissue as they provided positive detection in others assays (Figs 1-3). Antibody HPA046700 (Sigma) was also included in the analysis as it was shown to detect high levels of IRF5 in both red and white pulp of spleen by IHC from The Human Protein Atlas (http://www.proteinatlas.org/ENSG00000128604-IRF5/tissue/spleen). All four antibodies were tested by IHC and IF but only ab124792 and ab181553 worked for IF ( Fig. 4 and data not shown). Antibody cs3257 was unable to detect IRF5 by IHC or IF as staining was equivalent to background levels detected with secondary antibody alone ( Fig. 4B and data not shown). Ab124792 and ab181553 showed positive staining for IRF5, as compared to secondary alone, and the staining pattern was similar between the two antibodies. Both antibodies detected IRF5 in the red and white pulp, yet staining was distinctly higher in the red pulp. HPA046700 also showed positive staining; however, the pattern of staining was distinct from the two Abcam antibodies. Similar to that shown on the Human Protein Atlas website, high levels of IRF5 were detected in both red and white pulp of human spleen, with higher levels detected in the white pulp. White pulp is primarily made up of B and T cells. Given that IRF5 expression is extremely low in normal, healthy T cells, we performed co-staining with anti-CD3 antibodies to delineate antibody specificity. Data from IHC clearly show that ab124792 stained positive for IRF5 (brown) in cells lacking CD3 co-stain (red). However, HPA046700 stained positive for IRF5 in CD3-positive T cells suggesting non-specific binding by this antibody. Further analysis of co-staining by IF confirmed that ab124792 and ab181553 show IRF5-positive staining in cells that lack CD3 co-stain (Fig. 4D). Together, these data support the use of ab124792 or ab181553 for the specific detection of human IRF5 in FFPE samples by IHC or IF.   Last, we evaluated commercially available antibodies reported by the manufacturer to detect murine IRF5. For this study, spleen from wild-type (WT) and Irf5ko (KO) BALB/c mice was used to examine antibody specificity. Six antibodies were tested by immunoblot analysis and two showed highly specific detection of endogenous IRF5 -ab181533 (Abcam) and cs4950 (Cell Signaling); apparent by the single band of correct size in the WT lane and not in the KO lane (Fig. 5A). Ab33478 (Abcam) also detected a band in WT cells that corresponded to the size of IRF5 and was lacking in KO cells, yet numerous other non-specific bands were detected as well. Antibodies PT10547-1-AP (Proteintech), IC8447P (R&D), and SC98651 (Santa Cruz) were unable to detect differences in IRF5 expression between WT and KO splenocytes. Antibodies were then evaluated for their use in flow cytometry using manufacturer's recommended protocols for fixation, permeabilization and staining. Methods were further optimized in an identical manner by increasing Triton X-100 concentration and incubation time to enhance permeabilization. Data in Fig. 5B was from permeabilization of cells with 0.5% Triton X-100 for 20 min, followed by washing with 0.1% Triton X-100. All un-conjugated antibodies were tested with Alexa Fluor (AF) 488 and 647 secondary antibodies. Neither cs4950 nor PT105471AP were able to detect differential expression between WT and KO splenocytes. We also purchased two recently available pre-conjugated antibodies, PE-conjugated IC8447P (R&D; RDIC8447P-PE) and AF647-conjugated FAB8447R (Novus Biologicals; FAB8447R-AF647), for testing. Although a small shift between WT and KO cells was seen with each pre-conjugated antibody, MFI values were not substantially different between peaks (Fig. 5B). In our hands, only ab181553 was able to detect substantial differences in IRF5 expression between WT and KO splenocytes; MFI for WT was > 2-fold higher than KO. Antibodies were then tested for specificity by IF and IHC. Similar to findings from flow cytometry analysis, only ab181553 was able to differentiate IRF5 staining between WT and KO mice; both IF and IHC showed similar specificity. Antibody cs4950 was unable to detect any signal over that seen with secondary antibody alone, while PT105471AP gave positive staining that was not distinct between WT and KO spleen. Thus, ab181553 was the most specific in detecting murine IRF5 cellular expression by all three methods tested.

Discussion
Antibodies are among the most commonly used tools in the biological sciences to identify and isolate proteins of interest. Unfortunately, as documented in two seminal features in Nature last year 40,41 , commercial antibodies are "littering the field with false findings" 41 . The current study was an effort to validate commercial antibodies in a transparent manner with the inclusion of appropriate positive and negative controls for analysis. The underlying goal(s) of this study was to identify highly specific antibodies for the detection of human and murine IRF5, and to provide validated methods for testing and use of individual IRF5 antibodies to the scientific community. In the past 5 years, IRF5 has become a molecule of high interest given its important role(s) in regulating inflammation and immunity 7,8,10,19,22,24 , as well as recent new findings that implicate IRF5 in the regulation of neuropathic pain, obesity, myocardial infarction, allograft rejection, atherosclerosis, and metabolic dysfunction [25][26][27][28]42 . The body of research focused on IRF5 expression and function in disease pathogenesis continues to grow and therefore, like many other molecules of interest, will require the continual testing and validation of commercial reagents.
Not surprising to many scientists using antibodies in their research to detect and study molecules of interest, here we report that very few commercial antibodies developed for the detection of IRF5 were capable of detecting IRF5. To the contrary, many commonly used commercial antibodies were unable to accurately and/or specifically measure IRF5 protein expression. A cursory review of the literature indicates very few publications include validated positive and negative controls for protein expression analysis by immunoblot, flow cytometry or microscopy. These types of controls, including molecular weight markers, cell type-specific markers, and knockdown analysis, are essential for the accurate analysis of antibody specificity. In the current study, we identified ab181553 as the most specific antibody for detection of human IRF5 expression by immunoblot analysis (Fig. 1). For IP of human IRF5, NB100-1092, ab181553, and ab124792 were capable, yet ab124792 was the most specific for IRF5 (Fig. 2C-F). A number of antibodies showed positive staining (a shift to the right from isotype control) for human IRF5 by flow cytometry analysis, however, only cs13496, SAB1403991 and ab124792 were able to differentiate IRF5 expression levels between cell lines (Fig. 3). While not shown in the current study, pre-conjugated forms of ab124792 are now available and better at detecting intracellular IRF5 expression differences. Clone 2E3-1A11 used to be the only antibody available for the detection of human IRF5 by IHC and IF microscopy [20,43]. We now have identified two more antibodies that appear to specifically detect human IRF5 in FFPE tissues -ab124792 and ab181553 (Fig. 4). Further testing of HPA046700 is warranted given its current use in The Human Protein Atlas and our finding of IRF5 positive staining in splenic CD3-positive T cells with this antibody (Fig. 4C). The Human Protein Atlas is a public-domain resource used by many scientists for analysis of protein expression differences between normal and diseased cells or tissues. Last, we identified a single antibody that specifically recognizes murine IRF5 by immunoblot analysis, flow cytometry, and IHC/IF -ab181553. As shown in Figs 1 and 4, this antibody was also capable of specifically detecting human IRF5 by immunoblot analysis and IHC/IF.
A cursory review of immunogens used to generate each antibody may provide insight into the development of highly specific 2 nd generation α -IRF5 antibodies. For instance, the immunogens used to generate the previously validated rabbit polyclonal cs3257 and mouse monoclonal 2E3-1A11 were a synthetic peptide corresponding to carboxyl terminus residues of human IRF5 (cs3257) and full-length recombinant human IRF5 (2E3-1A11), respectively. Antibodies ab181553 and ab124792 are both rabbit monoclonal antibodies; ab181553 was generated from a recombinant fragment of amino terminal human IRF5 (amino acids 1-200) while the fragment used to generate ab124792 is proprietary. The immunogen for HPA046700 was a short region in the amino terminus of human IRF5 (amino acids 85-134). The amino terminus of IRF family members share significant homology as this is a well-conserved DNA binding domain. Conversely, the carboxyl terminus of IRFs is considered distinct from each other thus providing specificity of function 43 . While results reported herein identify α -IRF5 antibodies that specifically detect IRF5 in multiple assays (Table 1), caution should be used in the evaluation and interpretation of IRF5 expression analysis by other antibodies. Mice. Irf5 −/− mice on a C57Bl/6 background were obtained from T. Taniguchi (University of Tokyo, Tokyo, Japan) and T. Mak (University of Toronto, Toronto, Ontario, Canada) 8 . Wild-type (WT) Balb/c mice were purchased from The Jackson Laboratory (Bar Harbor, ME). Irf5 −/− mice on a C57Bl/6 background were back-crossed onto the Balb/c background to obtain a generation 10 (F10) cohort of Irf5 +/+ and Irf5 −/− Balb/c littermates, by standard breeding techniques. Mice were genotyped for Irf5 and the dedicator of cytokinesis 2 (Dock2) mutation as previously described 44,45 . All Irf5 +/+ and Irf5 −/− littermates used in this study lacked the Dock2 muta-tion45. Littermate Irf5 +/+ mice or WT Balb/c were used as controls. All animal experiments were conducted at the Feinstein Institute for Medical Research (FIMR), Northwell Health, following the guidelines of the Institutional Animal Ethics Committee. Animal ethics was approved by the FIMR Animal Ethics Committee and the experiments were carried out in accordance with the approved guidelines.

Methods
Real-time quantitative polymerase chain reaction (qRT-PCR) analysis. Primers for IRF5 amplification and thermocycler conditions were previously described 32 . All reactions were performed in the QuantStudio 3 Real-Time PCR System (ThermoFisher Scientific). PCR analyses were done in duplicate, with each set of assays repeated three times. To minimize the effects of unequal quantities of starting RNA and to eliminate potential sources of inconsistency, relative expression levels of each gene was normalized to β -actin using the ∆∆Ct method.
Flow cytometry. For human cell lines, 2 × 10 6 cells were fixed overnight at 4 °C in 2% formaldehyde, washed, and permeabilized in 0.1% Triton X-100 as previously described 33 Human, H; mouse, M. + + , antibody worked very well and was specific; + , antibody worked; + /− , antibody was able to detect IRF5 but not specific; NS, signal was detected but not specific for IRF5.