Purification of Mitochondrial Proteins HSP60 and ATP Synthase from Ascidian Eggs: Implications for Antibody Specificity

Use of antibodies is a cornerstone of biological studies and it is important to identify the recognized protein with certainty. Generally an antibody is considered specific if it labels a single band of the expected size in the tissue of interest, or has a strong affinity for the antigen produced in a heterologous system. The identity of the antibody target protein is rarely confirmed by purification and sequencing, however in many cases this may be necessary. In this study we sought to characterize the myoplasm, a mitochondria-rich domain present in eggs and segregated into tadpole muscle cells of ascidians (urochordates). The targeted proteins of two antibodies that label the myoplasm were purified using both classic immunoaffinity methods and a novel protein purification scheme based on sequential ion exchange chromatography followed by two-dimensional gel electrophoresis. Surprisingly, mass spectrometry sequencing revealed that in both cases the proteins recognized are unrelated to the original antigens. NN18, a monoclonal antibody which was raised against porcine spinal cord and recognizes the NF-M neurofilament subunit in vertebrates, in fact labels mitochondrial ATP synthase in the ascidian embryo. PMF-C13, an antibody we raised to and purified against PmMRF, which is the MyoD homolog of the ascidian Phallusia mammillata, in fact recognizes mitochondrial HSP60. High resolution immunolabeling on whole embryos and isolated cortices demonstrates localization to the inner mitochondrial membrane for both ATP synthase and HSP60. We discuss the general implications of our results for antibody specificity and the verification methods which can be used to determine unequivocally an antibody's target.


Introduction
Antibody tools are of fundamental importance for learning about proteins and cellular mechanisms, and a number of recent reviews are calling for rigorous standards of antibody validation [1][2][3][4][5]. Concern about antibody specificity is increasing with the use of fluorescent protein fusions to determine protein distribution and the need to resolve discrepancies between live and fixed samples [6]. Frequently in the literature the specificity of an antibody is considered verified if it labels a single band of the expected size, or shows an affinity for the antigen produced in a heterologous system such as bacteria. Here we challenge these assumptions by directly purifying two antibody targets from the ascidian egg using both immunoaffinity methods and a novel strategy to enrich for nonabundant proteins.
We sought to characterize the polarized mitochondria-rich domain of ascidian eggs known as the myoplasm. Ascidians are tunicates, the sister group to vertebrates [7], and their embryos develop rapidly into tadpole larvae exhibiting the typical chordate body plan [8,9]. Over a century ago observations of the inheritance of myoplasm led biologists to propose the theory of ''mosaic'' or autonomous development whereby localized domains containing organelles and molecular determinants are partitioned into distinct cell lineages [10,11]. The myoplasm is positioned as a peripheral basket in the ascidian egg, and after fertilization it moves via a series of cytoskeletal reorganizations to form a posterior crescent shape which will segregate into tail muscle [12][13][14] (Fig. 1). In living cells, the myoplasm is easily visible due to differential pigmentation, autofluorescence or via the application of vital mitochondrial dyes [10,13,[15][16][17][18][19][20]. The eggs of many animals contain aggregates of tightly packed mitochondria, often associated with germ plasm, which migrate together in a mitochondrial cloud, or Balbiani body [21][22][23]. The ascidian myoplasm offers an accessible system to address questions concerning networks of mitochondria, such as their regulation, function, and how they are properly partitioned during cell division. Ascidians are amenable to an increasing battery of experimental approaches including micromanipulation, modification of gene expression or function, live imaging, genetics, proteomics [20,[24][25][26][27][28] and as we show here, biochemistry.
The most commonly used antibody tool to study ascidian myoplasm is ''NN18'', a monoclonal originally made to a neurofilament preparation from pig spinal cord [29]. NN18, also called NF-160 or NF-M, reacts with the medium molecular weight (160 kDa) neurofilament subunit and labels exclusively neurofilaments in vertebrates as well as crab [30][31][32][33]. In the ascidian, it was found that NN18 recognizes a 58 kDa protein and strongly labels myoplasm in eggs from numerous species [13,[34][35][36][37]. The ascidian target of NN18 known as p58 interacts with ''myoplasmin'' whose sequence has features characteristic of proteins which form filaments [36,38]. Since the myoplasm can be isolated as a unified mass [37,39,40] and early electron microscopy studies suggested that it contains structures resembling intermediate filaments [41][42][43], it was postulated that the ascidian target of NN18 is an intermediate filament-like protein as in vertebrates. However a more recent analysis of the Ciona intestinalis genome showed that ascidians lack the neurofilament (type IV) class of intermediate filaments [44], so the identity of ascidian p58 remains unclear.
Another unresolved question concerns the function of the myoplasm: it is thought to provide abundant energy for contracting myofibrils of the tadpole tail but the original proposition that it harbors molecules involved in myogenesis has not been ruled out. The myogenic determinant Macho RNA [45,46] is localized to a domain rich in endoplasmic reticulum positioned just adjacent to the myoplasm, known as the CAB (see Fig. 1) [47][48][49][50][51][52][53]. Macho protein leads to activation of zygotic expression of the muscle-specific transcription factor MyoD in the muscle lineage [54][55][56]. Ascidian oocytes contain a low level of maternal mRNA encoding MyoD [57,58] however whether a corresponding maternal MyoD protein is present in the egg or myoplasm has not been addressed.
Here we generate a polyclonal antibody against ascidian MyoD, show it labels the myoplasm, and isolate its target from ascidian eggs. We also identify with certainty the target of the standard myoplasm marker antibody NN18. Our unexpected findings have general implications for antibody ''specificity'' and highlight the necessity for unequivocal validation of antibody tools.

Materials and Methods
For additional methods and details, see Methods S1.

Animals
Adults of Phallusia mammillata were collected in the Mediterranean near Sete, France. All necessary permits were obtained for the described field studies from the Minister of Ecology and Sustainable Development, Marseille. The European ascidian Phallusia is favorable for biochemical approaches and protein purification because it produces large quantities of eggs (up to 1 ml .10 6 eggs per hermaphroditic animal). Mature eggs are less abundant in Ciona intestinalis oviducts, however the Ciona gonad is an excellent source of maternal protein for biochemistry. This large bean-shaped organ can be up to 1 cm long and contains a virtually pure population of unfertilized eggs and immature oocytes of all stages [59]. Ciona adults were obtained from Roscoff marine station (Brittany, France) or through the National Bio-Resource Project NBRP of MEXT, Japan. For general ascidian protocols, fertilization and culture of embryos, see [20].

Primary antibodies
The NN18-clone, originally generated as one of a panel of monoclonal antibodies to porcine spinal cord [29], was purchased from Sigma-Aldrich or ICN (anti-neurofilament 160 mouse monoclonal). Antibody PMF-C13 was produced by injecting a peptide present in PmMRF which is the Phallusia homolog of MyoD (Genbank accession HQ287931) into rabbits; immune serum was affinity purified against bacterially expressed PmMRF (Fig. S1).

Enrichment for proteins with low affinity for ion exchange resin
In initial 2D gels loaded with total protein extract from Phallusia eggs or Ciona ovary, p63 was not abundant enough to be visible by coomassie staining, therefore a sample highly enriched for p63 was prepared by sequential ion exchange chromatography. The majority of egg proteins were eliminated by binding to a saturated DEAE column, and the flowthrough was applied to a second column on which p63 was retained. Fractions were eluted via a step salt gradient and assayed for p63 by immunoblot with PMF-C13; positive fractions were concentrated and loaded onto preparative 2D gels. See Methods S1 for details.
In order to address the question of maternal MyoD protein, we generated a polyclonal antibody using as antigen a 13 amino acid peptide sequence present in the Phallusia MyoD homolog PmMRF (Fig. 3A). This antibody, named PMF-C13, passed tests commonly cited as adequate to verify specificity for its antigen: it specifically recognizes the antigen protein PmMRF produced in bacteria, but not truncated forms of PmMRF which lack the antigenic peptide ( Fig. 3B), and in ascidian extracts it strongly labels a single band of the expected size which we denote p63 (Fig. 3C,D). Like p58 ( Fig. 2A), p63 is maternally provided and stably present at all embryonic stages in both Phallusia and Ciona (Fig. 3C, D). Immunofluorescence using PMF-C13 purified by affinity chromatography against PmMRF (Fig. S1) showed strong and persistent labeling of the ascidian myoplasm throughout development of Phallusia or Ciona (Fig. 3E), displaying a distribution very similar to that obtained with monoclonal NN18 (Fig. 2B). This localization pattern is surprising, since PmMRF is a myogenic transcription factor expected to be found in the nuclei of muscle cells not in association with a mitochondria-rich domain.
Thus the identity of p63, like that of p58, remained in question and has important implications for the structure and function of the ascidian myoplasm. In order to determine whether the myoplasm contains a neurofilament-like or MyoD-like protein, we set out to identify with certainty the proteins recognized by NN18 and PMF-C13, p58 and p63. Purification of p58 and identification as ATP synthase The identification of p58 was accomplished by 2 independent methods: immuno-precipitation and immunoscreening. Sepharose resin coupled to NN18 antibody was incubated with a soluble fraction from Ciona gonad homogenate and bound proteins were separated on a non-reducing SDS-PAGE gel, so that the intact NN18 IgG would migrate at a high molecular weight well separated from the target protein. p58 was successfully immunoprecipitated as determined by immunoblot (Fig. 4A, ''western''), and a duplicate lane silver-stained for total protein (Fig. 4B, right) showed that it was highly purified and abundant enough to obtain N-terminal sequence. The resulting sequence of 27 amino acid residues (circled in Fig. 4B) is identical in 26 positions to residues 44-70 of Ciona mitochondrial-type ATP synthase alpha-subunit. As this was an unexpected result, we also screened a CionacDNA expression library with NN18 antibody. All positive clones encoded the same protein: ATP synthase alpha-subunit ''CiATP-synA'' (''Immunoscreen'' line, Fig. 4B). The first 43 amino acids of Ciona ATP synthase missing from the immunoprecipitated p58 correspond to the transit signal peptide (boxed in Fig. 4B and Fig.  S2C) which targets proteins to mitochondria and is cleaved during the import process. Thus p58 protein recognized by the neurofilament NN18 antibody in the Ciona egg is the mature form of ATPsynthase alpha subunit. This protein along with ATP synthase beta subunit make up the intramitochondrial F1 portion of the enzyme complex which is attached to the F0 portion embedded in the mitochondrial membrane. This cross reaction is somewhat of a mystery, as there are no significant stretches of homology between ascidian ATP synthase alpha (Fig. S2B) and the initial antigenic protein, porcine NeuroFilament-M [29] (Fig.  S2A), but it remains possible that they possess a short stretch of identical amino acids or a resemblance in tertiary structure.

Purification of p63 and identification as HSP60
Attempts at immunoprecipitation and immunoscreening with PMF-C13 were unsuccessful, and in such cases the target protein must be purified biochemically. p63 was isolated from Phallusia eggs by a combination of ion exchange chromatography and isoelectric focusing (Fig. 5). Initial immunoblots of 2D gels covering a range of pH revealed that the isoelectric point of p63 was approximately 5.2 but that it was not abundant enough to be distinguished as a major spot within the constellation of total egg proteins. An enrichment of p63 was obtained by a strategy of 2 sequential ion exchange columns (schematized in Fig. 5B). Initial DEAE columns indicated a relatively weak affinity for anionic exchange DEAE resin (Fig. 5A). A first DEAE column was intentionally overloaded with saturating amount of Phallusia egg extract (Fig. 5C), selecting for binding of proteins with stronger affinity for DEAE. Under these conditions the majority of proteins (over 80%) were retained, but p63 flowed through the column with a minority of proteins having weaker affinity for DEAE (Fig. 5C  ''column 1'' FT). The flow-through from this first column which was poor in protein content but enriched for p63 was used to load a second non-saturated ion exchange column, on which p63 was retained (Fig. 5C ''column 2''). The fractions eluted with 200 mM NaCl highly enriched for p63 were concentrated and loaded on two identical isoelectric focusing gels followed by PAGE.
One preparative 2D gel was stained with Coomassie blue and the twin gel was used for western blot with antibody PMF-C13 (Fig. 5D). A precise overlay of the blot and gel (see Methods S1) revealed that p63 was sufficiently abundant and pure to be excisedand subjected to mass spectrometry analysis. Peptide sequences obtained were compared to databases of the translated Ciona genome and subsequently to that of Phallusia ESTs. In both cases the strongest match was mitochondrial Heat Shock Protein 60 (HSP60) (Fig. S3A), which is a molecular chaperone of the GroEL family required for protein folding and mitochondrial activity. Ascidian HSP60 shares a similar molecular weight and isoelectric point with the antigen protein PmMRF (Fig. S3B,  legend), but there is no obvious sequence or structural similarity between the two to explain the affinity of antibody PMF-C13. Nor could this target protein have been predicted: a search of Ciona gene models shows that 5 proteins contain a stretch of 6 consecutive amino acids in common with the antigenic peptide CTDALNEQLSMLQ, and over 200 Ciona proteins contain a 5 amino acid stretch, however the bona fide antibody target HSP60 is not on that list.

ATP synthase and HSP60 are located on the inner mitochondrial membrane
Thus both the NN18 and PMF-C13 antibodies recognize not the original antigens, but instead two well conserved proteins predicted to localize in mitochondria.
The distribution of ATP synthase and HSP60 in whole eggs and embryos was examined by high resolution confocal microscopy (Fig. 6). In the unfertilized ascidian egg three peripheral regions can be distinguished: the vegetal layer containing the myoplasm basket rich in mitochondria ( Fig. 6E and H), the animal  hemisphere relatively poor in mitochondria ( Fig. 6C and F), and a transition zone at the equator ( Fig. 6D and G). Both NN18 (red, Fig. 6C-E) and PMF-C13 (green, Fig. 6F-H) recognize the same individual mitochondria in every region of the egg (yellow overlay, arrows in Fig. 6). Absence of signal corresponded to cell regions lacking mitochondria such as the animal pole (Fig. 6C ''ap''), nuclei (Fig. S4 ''n'') and the CAB (arrows in Fig. 3E, see Fig. 1). While a previous study indicated that HSP60 is found preferentially in myoplasm mitochondria [60], we observe that mitochondria in all cells of embryos and tadpoles are labelled (Fig. S4) and conclude that the myoplasm staining is due to the massive enrichment of mitochondria in this domain. In addition PMF-C13 reproducibly labeled smaller non-mitochondrial spots (arrowheads Fig. 6 F-H) distributed throughout most of the cytoplasm. NN18 labelled exclusively mitochondria in these preparations, however the use of sub-cellular fractionation followed by immunnoblotting indicates that p58 is also present in non-mitochondrial cytoplasm [61].
In order to confirm mitochondrial association we used isolated cortices (Fig. 7), which are open-cell preparations of plasma membrane and adherent organelles [62,63]. We have noted that cortices prepared from cleaving ascidian eggs retain patches of myoplasm, yielding a coverslip of semi-purified mitochondria. Such cortical preparations were first labeled with DiO(C 6 )3 to identify mitochondria, then fixed and immunolabled with antibodies. On cortices treated with triton in order to expose the inner mitochondrial membrane, NN18 and PMF-C13 antibodies label spots which colocalize with each other (Fig. 7A, B, D). In the absence of permeabilization with detergent however, antibodies only have access to the outer surface of mitochondria and NN18 and PMF-C13 do not label the isolated cortex ( Fig. 7C and E, right panels). These results demonstrate that ATP synthase and HSP60 are present within mitochondria in ascidian eggs. Other subcellular localizations, such as the granules observed in Fig. 6 F-H, would not be detectable by this method however since during cortex preparation, cytosol and structures not adherent to the plasma membrane are eliminated.

Discussion
In this study the application of biochemical methods to ascidian eggs allowed us to identify unambiguously two proteins recognized by myoplasm antibodies. Somewhat alarmingly, in both cases the antibody's target was unrelated to the original antigenic protein.
The polyclonal antibody PMF-C13 showed a strong affinity for its intended target PmMRF (Figs. 3, 5, S1) and yet labels HSP60 (Figs. 5, S3). Equally unexpectedly, we find that the monoclonal antibody NN18 recognizes a protein wholly unrelated to the intermediate filament antigen, ascidian ATP synthase subunit alpha (Figs. 2, 4, S2). Thus a significant message of the present work is that use of unvalidated antibodies can lead to erroneous conclusions, such as in our case that the myoplasm is structured by intermediate filaments, or that the MyoD transcription factor localizes to mitochondria.
While the causes of these cross-reactions are unknown, their occurrence may be representative of general situations which lead to mistaken identity, some of which have been documented [4,6,64,65]. For instance a secondary affinity for an abundant protein may be revealed when the presumed antigen is not present in the tissue of interest, as is the case for the NN18 antibody. The Ciona genome encodes 5 intermediate filament genes [44] but no homolog to the neurofilamant (typeIV) class which antibody NN18 recognizes in vertebrates. ATP synthase is one of the most abundant proteins in eggs as determined by analysis of the Ciona proteome (http://cipro.ibio.jp/2.5/expression profile KH.C10.579 [66]. The constant level of ATP synthase throughout embryonic development ( Fig. 2A and [34] likely reflects stable maternal protein, since we find that the ATP synthase transcript is abundant in the egg and decreases during cleavage stages (Fig. S5), a pattern like that of the transcript encoding ascidian ATP translocase [67]. It is worth pointing out that since p58/ATP synthase is so stable and well-characterized, the commercially available NN18 monoclonal can be used by the ascidian community as a reliable standard to compare protein amount among samples.
In the case of the polyclonal PMF-C13, we found it recognizes HSP60 (Fig. 5, S3) even though it passed tests commonly cited to assert antigen specificity, namely that it labels a single band of the expected size and isoelectric point in the tissue of interest, and it recognizes the antigen (PmMRF) produced in a heterologous system but not truncated forms lacking the peptide sequence (Figs. 3, 5, S1). HSP60 contains 3 short sequences in common with the antigenic peptide (DAL, LNE, DALN, shaded in Fig. S3C) which could function as epitope. Indeed, since any short series of amino acids will be fortuitously present in numerous polypeptide chains, such a misleading binding of antibodies to unexpected targets may be relatively common. It is possible that antibody PMF-C13 also recognizes PmMRF in ascidian extracts, but given their similar molecular weights (Fig. S3 legend), the band corresponding to this transiently expressed transcription factor would be obscured by the stronger signal of HSP60. HSP60 is primarily a mitochondrial protein, but it is also found the cytosol and on the cell surface where it is implicated in diverse cellular functions including protein trafficking through the plasma membrane, cell signaling, apoptosis and immunological response [68][69][70][71][72][73][74]. The non-mitochondrial spots labeled by PMF-C13 (arrowheads in Fig. 6 F-H) are reminiscent of the HSP60containing cytoplasmic granules of unknown function observed in mammalian cells [68]. HSP60 binds the translation factor YB-1 in the cytoplasm of mammalian cells [72] suggesting that the cytoplasmic HSP60 may be part of a polysome RNP complex. Interestingly in Ciona YB-1 binds to and regulates translation of maternal mRNAs including the localized determinants macho and PEM [75]. The ascidian and in particular the transparent Phallusia embryo which is favorable for live imaging [13,20,25] is a promising model for investigating the multiple locations and functions of HSP60 and of ATP synthase.
What methods are available for definitive validation of antibody tools and when should they be applied? It is well known that fixation conditions can radically alter the perceived distribution of a protein [6,[76][77][78] thus leading to potential misinterpretation about a protein's bona fide localization or function. To avoid fixation artifacts and to palliate the lack of antibodies it is common to infer protein localization by following fluorescently-tagged protein fusions in living cells, however an exogenous overexpressed modified protein may not faithfully reflect the distribution of the endogenous protein. Thus in general, when the distribution of a protein determined by immunolabelling is novel, unexpected, or unlike that of the GFP-tagged version, it is necessary to verify the antibody target in the species under study. In genetic model systems, homologous gene replacement or mutant collections can be employed to demonstrate that an altered coding sequence leads to the expected change in size or localization of the recognized protein. In many species and cell types one can use RNA interference or morpholino oligonucleotides to inhibit translation of a specific RNA transcript and show that there is a corresponding reduction of the candidate band or immunofluorescence signal. However proteins which are very stable and/or are provided maternally will be little affected by this type of translational inhibition. Moreover these methods are indirect: knockdown of factors involved in transcription or signaling pathways for example may also reduce the levels of their targets. Ultimately the most direct way to determine an antibody's target is to isolate it from the material of interest by the methods described in the present article: immunoprecipitation, immunoscreening of an expression library, or biochemical purification. This latter approach should become less onerous with the development of large scale proteomics, whereby researchers or commercial companies can pre-verify uncharacterized antibodies. aligned peptides showed a match to the same protein HSP60, with a higher percentage of sequence identity (shaded) to Phallusia mammillata, the species from which p63 was isolated, than to Ciona. Ciona HSP60 (gene model KH.C6.85 on Ciona genome browser http://ghost.zool.kyoto-u.ac.jp/SearchGenome kh.html [1] is a protein of 573 amino acids with theoretical molecular weight 61.4 kD and isoelectric point 5.36. Phallusia HSP60 is a protein of 577 amino acids with theoretical molecular weight 61.9 kD and isoelectric point 5.40. The PmHSP60 sequence was compiled by manual assembly of unpublished Phallusia EST data on the bioinformatics server ''Octopus'' at Villefranche. This is a first demonstration that the well-developed Ciona proteomics database [2] can be used to identify proteins from the related species Phallusia mammillata.  Methods S1 (DOCX)