Authentication of collagen VI antibodies

A major discussion point within the biomedical research community concerns experimental reproducibility. One recent study attempted to replicate the findings from 53 clinical experiments but found that results from just six of these studies could be independently confirmed [1]. While there are many reasons for this, a primary factor is the inadequate validation of antibodies leading to false positive (and negative) findings. Many antibodies in the public domain have been found to be poorly characterized. For example, a 2008 study found that only half of >6000 commercially-available antibodies uniquely recognized their stated targets [2]. The most common issue was poor specificity, where the antibody recognizes structurally-similar proteins in addition to the target protein. As the use of antibodies in research increases, it is critical that proper antibody authentication experiments are conducted. Funding bodies are also concerned about the apparent lack of experimental reproducibility. In the US, National Institutes of Health requires a validation plan to be provided for key research tools, including antibodies, in all grant applications from 2016 onwards (see http://grants.nih.gov/reproducibility/index.htm).

Collagen VI is present in the extracellular matrix (ECM) of skeletal muscle where it functions to anchor the basement membrane to underlying interstitial tissues. Five collagen VI chains have been described in human [3, 4]. Mutations in three of these (COL6A1, COL6A2 and COL6A3) result in two congenital muscular dystrophies demonstrating that collagen VI is critical for muscle development. These diseases are the relatively mild Bethlem myopathy (BM) and the more severe Ullrich congenital muscular dystrophy (UCMD) [5, 6]. Studies on skeletal muscle biopsies and cultured fibroblasts isolated from affected individuals have established that, in general, the severity of disease is determined by the type of mutation. The majority of disease-causing mutations are amino acid substitutions that typically result in reduced efficiency of collagen VI assembly, leading to reduced and/or aberrant deposition of collagen VI in the muscle ECM [7–10]. Antibodies are used to assess the consequences of mutant or absent protein on collagen VI assembly by immunoblotting, and on ECM deposition and organization by immunohistochemistry [9, 11–14]. Commonly-used antibodies include VI-26 (MAB3303) raised against tissue-purified collagen VI, 3C4/Mab1944 for the α3 chain, and 6A1-H200 for α1. Public discussion within the collagen VI community about the use of antibodies for the molecular and histological analysis of collagen VI mutations [15, 16] stimulated an effort to authenticate commonly-used antibodies. In this report we define chain specificity and relative sensitivities of 23 commercially-available collagen VI antibodies by immunohistochemistry and immunoblotting.

To assess the relative efficiency of collagen VI antibodies to detect collagen VI in muscle, cross and longitudinal frozen sections of normal human skeletal muscle were immunostained with a panel of 23 commercially-available collagen VI antibodies. Primary antibodies were detected by fluorescently-labelled secondary antibodies. Experimental conditions were identical except for antibody dilutions, which were selected based on manufacturers guidelines, and choice of secondary antibody which was dictated by the species the primary antibody was raised in.

Chain specificity was also confirmed for three antibodies by immunocytochemistry on transfected cells (Additional file 1: Figure S1). These antibodies were; ab199720 for α1, ab180855 for α2 and 70R-CR009X for α3. While these cells do not produce a collagen VI extracellular matrix because they lack the other chains necessary for heterotrimeric assembly, they are used in the current study to confirm chain specificity as indicated by immunoblot data.

Antibodies raised against tissue-purified collagen VI

Five antibodies that were raised against collagen VI from pepsin-treated tissue were tested (Fig. 1). Of these, only two showed detectable signals by immunohistochemistry: monoclonal MAB3303 (clone VI-26 from Millipore) (panels A and B) and polyclonal 70R-CR009X (Fitzgerald Industries) (E and F). The staining was clearly pericellular in cross- and longitudinal sections, as expected for collagen VI. In addition, both antibodies recognized high molecular weight collagen VI in lysates from cells transfected with all three chains (D and H). Interestingly, the Fitzgerald Industries antibody also recognized individual α3 chains when electrophoresed under non-reducing conditions suggesting that the α3 chain may be the major epitope for this antibody (G). This finding was confirmed in transfected cells where 70R-CR009X stained cells expressing α3 chains but not those expressing α1 nor α2 chains (Additional file 1: Figure S1, panels C, H and M). This is surprising since native collagen VI was used as the immunogen and may suggest that α3 chains are ‘preferentially’ exposed on native collagen VI fibrils. Alternatively, the α3 chain may have a more immunogenic structure than α1 and α2 chains in native collagen VI. Of the antibodies that failed to stain tissue, 5C6 is of interest because it has been used previously in immuno-gold labeling experiments to detect collagen VI in skin and cartilage [20, 21] where presumably collagen VI structures are in their native form. It is unclear why 5C6 failed to stain muscle frozen tissue (M and N) but the 5C6 epitope may not be revealed by the fixation and section preparation used in this study, or alternatively, skin and cartilage collagen VI has a different supramolecular architecture than muscle collagen VI and epitope accessibility varies between the two tissues. The sc-47764 antibody (clone 172C2 from Santa Cruz) did not detect muscle collagen VI by immunohistochemistry (I and J) but recognized individual α3 chains by blotting (K). The Millipore antibody AB7821 did not detect collagen VI by immunohistochemistry and reacted with multiple bands by immunoblotting (Q-T).

Antibodies directed against individual collagen VI chains

α1(VI)

Four of the five antibodies against the α1 chain produced a signal on skeletal muscle sections, with ab199720 (abcam) (Fig. 2, panels E, F) and 17023-1-AP (ProteinTech Group) (M and N) detecting highest levels in the pericellular space. The sc-20649 (clone H200 from Santa Cruz) staining intensity was weaker and intracellular structures appeared to be positive (A and B). Chain specificity immunoblots revealed that all five anti- α1 antibodies tested recognized α1 chains, and no other chains, when resolved under reducing conditions at the predicted molecular weight of 140 kDa (C, G, K, O and S). Antibodies 17023-1-AP and ab199720 recognized α1 chains in both reduced and non-reduced samples (G and O). Interestingly, both these antibodies revealed the presence of discrete α1 oligomers larger than 140 kDa suggesting that α1-chains may undergo self-assembly to form homo-oligomers in these cells. Four antibodies recognized α1 chains at approximately 180 kDa in lysate containing all three chains which corresponds to the size of α1 dimers (D, H, L and P). All α1 antibodies recognize a small amount of high-molecular weight collagen VI near the top of the blots.

In transfected cells, ab199720 gave a strong signal in cultures of cells expressing α1, but not in cells expressing α2 or α3 chains (Additional file 1: Figure S1, panels A, F and K). This confirms the immunoblot finding (see Fig. 2G) that this antibody specifically recognizes the α1 chain.

α2(VI)

Six out of seven α2-antisera stained skeletal muscle, with ab180855 (abcam) demonstrating the strongest signal (Fig. 3, panel D). Only sc-374566 (clone B7 from Santa Cruz) failed to detect muscle collagen VI (W and X). All seven α2 antibodies recognized their epitopes when lysates were resolved under reducing conditions although with a range of sensitivities. Three α2 antisera worked well (F, M and Y) with clearly detectable bands at the correct molecular weight (140 kDa) and four performed poorly with only faint α2 bands present (C, J, Q and U). Only sc-374566 recognized individual non-reduced α2 chains, albeit weakly (Y). Several antibodies recognized multimeric collagen VI with sc-374566, sc-83607 (clone D20 from Santa Cruz), and sc-377143 (clone K15 from Santa Cruz) (asterisks in panels N, V and Z) showing the strongest signals. ab180855 stained cells expressing α2 but not α1 nor α3 (Additional file 1: Figure S1, panels B, G and L) confirming immunoblot findings for this antibody (Fig. 3F).

α3(VI)

Three of six α3 antibodies stained skeletal muscle including the commonly used 3C4/MAB1944 (Millipore) which also appeared to stain intracellular structures (Fig. 4 panels A, B, E, F, I and J). 3C4/MAB1944 and sc-131139 (clone G18) recognized their 260 kDa target under non-reducing conditions [22, 23] (C and G). The remaining α3 antibodies failed to recognize trimeric collagen VI nor individual chains under reducing or non-reducing conditions. Three antisera; sc-81766 (clone H3-2 from Santa Cruz), ab49273 (clone 64C11 from Abcam) and sc-131140 (clone N12 from Santa Cruz) failed to recognize collagen VI by either method using the experimental conditions used in this study (M to V).

In the course of these studies we became aware that the mouse monoclonal antibody 3C4/MAB1944, which recognizes the α3(VI) chain, gave different results depending on supplier and batch number (Fig. 5). As shown in Fig. 5C, 3C4 from Santa Cruz recognized α3(VI) at the correct molecular weight of 260 kDa when resolved under reducing conditions. In contrast, the same antibody clone from Millipore recognized α3 at 260 kDa and additional bands at 50, 70 and 90 kDa depending on lot number (Fig. 5A, B). 3C4 is widely used in studies on human muscle using immunohistochemistry. Our data suggest that care should be taken in interpreting data using these batches of antibody from Millipore.

Controls

Control experiments where sections were stained using rabbit, mouse and goat IgGs at dilutions of 1:100 did not detect any collagen VI (Additional file 2: Figure S2, panels A–F). Similarly, blots containing lysates from triple transfected cells probed with mouse IgG and normal rabbit and goat serum failed to detect significant bands (G–I). A faint band at 70 kDa is present on the blot probed with normal rabbit serum but this is smaller than the 140 kDa α1 and α2 bands and 260 kDa α3 bands and represents background (asterisk in panel H).

To provide a meaningful assessment of the relative usefulness of collagen VI antibodies for immunohistochemistry 23 collagen VI antibodies from Abcam, Santa Cruz, ProteinTech Group and Millipore were tested for reactivity on human muscle sections under standard conditions. Since aberrant collagen VI tissue deposition can be caused by mutations in three genetically-distinct collagen VI chains and it would be valuable to know whether any antibody cross-reacted with the other structurally-similar chains, each antibody was tested for chain specificity by immunoblot analyses and, for three antibodies, immunocytochemistry. Comparative testing revealed that 15 out of 23 antibodies detected pericellular collagen VI on human skeletal muscle sections. The best antibodies specific for each chain are: ab199720 and 17023-1-AP for α1, ab180855 and 14853-1-AP for α2, 3C4/MAB1944 and sc-367543 for α3 (see Table 2). The only two antibodies raised against tissue collagen VI that detected multimeric collagen VI by immunocytochemistry were MAB3033 and 70R-CR009x.

14 out of 18 chain-specific antibodies recognized their target collagen VI chains although with a wide range of sensitivities. Notably, all 14 antibodies demonstrated chain specificity with each antisera recognizing a single collagen VI chain and none appeared to cross-react with the other collagen VI chains. However, good specificity within the collagen VI family does not exclude the possibility that these antibodies bind to other collagens or non-collagenous proteins. Each distinct chain was recognized by at least two different antibodies with high sensitivity (Table 2). The finding that antisera against α1 and α2 primarily recognizes the reduced form of the chain suggests that these chains form intra- or inter-chain disulfide bonds that nominally mask the epitopes, and complete reduction of disulfides is essential for optimal antibody recognition. Epitopes within high molecular weight collagen VI were detected by multiple antibodies with the three strongest signals being ab182744 against α1, and sc-374566 and sc-83607 against α2.

There has been discussion in the collagen VI literature regarding the use of VI-26/MAB3303 and 3C4 for immunostaining patient muscle tissue and fibroblasts [15, 16]. It has been reported that VI-26 is less sensitive at detecting collagen VI in muscle and fibroblast preparations compared to 3C4 leading to an underestimation of the amount of collagen VI degradation [23]. Our data indicates that they recognize different forms of collagen VI; VI-26 detects high-molecular weight, (presumably) triple helical collagen VI but not any individual chains (Fig. 1C, D) whereas 3C4 detects the α3 chain and not collagen VI multimers (Fig. 4G, H). These antibodies clearly recognize different epitopes and may detect different sub-populations of collagen VI fibrils in the ECM. We suggest staining using more than one antibody when detecting collagen VI by immunohistochemistry.

This study expands the list of collagen VI antibodies that can be used in immunoblotting for the analysis of collagen VI on patient samples. Antibodies with high specificity and sensitivity are available for each collagen VI chain. Most α1 and α2 antisera perform best on reduced material; in contrast the majority of α3 antisera tested performed better on non-reduced samples. Several antibodies recognize multimeric collagen VI with the strongest signal from the 70R-CR009X antibody.

There have been calls to standardize antibody production, characterization and use [24, 25]. It is recommended that journals insist that authors be more transparent and consistent with reporting antibody details [26]. We echo this sentiment and reiterate the need to include basic information when reporting antibody experiments by including dilutions used, vendor and catalogue/clone numbers, and species antibody was raised in. In addition, it is essential to include controls such as isotype controls for monoclonal antibodies, pre-immune serum controls for polyclonal antisera and blocking peptide controls for peptide antibodies, if possible.