1. Cell Biology
  2. Immunology and Inflammation
Download icon

A two-hybrid antibody micropattern assay reveals specific in cis interactions of MHC I heavy chains at the cell surface

Research Article
  • Cited 0
  • Views 496
  • Annotations
Cite this article as: eLife 2018;7:e34150 doi: 10.7554/eLife.34150

Abstract

We demonstrate a two-hybrid assay based on antibody micropatterns to study protein-protein interactions at the cell surface of major histocompatibility complex class I (MHC I) proteins. Anti-tag and conformation-specific antibodies are used for individual capture of specific forms of MHC I proteins that allow for location- and conformation-specific analysis by fluorescence microscopy. The assay is used to study the in cis interactions of MHC I proteins at the cell surface under controlled conditions and to define the involved protein conformations. Our results show that homotypic in cis interactions occur exclusively between MHC I free heavy chains, and we identify the dissociation of the light chain from the MHC I protein complex as a condition for MHC I in cis interactions. The functional role of these MHC I protein-protein interactions at the cell surface needs further investigation. We propose future technical developments of our two-hybrid assay for further analysis of MHC I protein-protein interactions.

https://doi.org/10.7554/eLife.34150.001

Introduction

Protein-protein interactions are difficult to investigate, especially when they involve membrane proteins under physiological conditions, specific protein conformations or subpopulations, low affinities, or defined locations in the cell. Such challenges are not usually met by yeast two-hybrid screens and co-immunoprecipitation approaches; instead, they require technically demanding methods such as fluorescence resonance energy transfer (FRET) or fluorescence correlation spectroscopy, which only work in some cases.

Recently, antibody-based capture assays on solid supports have been described that can be used in bait-prey experiments in live cells (Löchte et al., 2014; Schwarzenbacher et al., 2008; Weghuber et al., 2010). We now demonstrate an expansion of this concept to characterize location- and conformation-specific protein-protein interactions in a novel two-hybrid assay read out by fluorescence microscopy by using microprinted antibody patterns for the capture of bait proteins to spatially arrange them in the plasma membrane of live cells (Dirscherl and Springer, 2017) and to investigate their interaction with green fluorescent protein (GFP)-tagged prey proteins. The assay is universally applicable for the investigation of protein-protein interactions.

In this paper, we use this two-hybrid micropattern assay to solve the long-standing question which forms of major histocompatibility complex class I (MHC I) proteins associate laterally (in cis) on the plasma membrane. MHC I proteins consist of a polymorphic transmembrane heavy chain (HC), the non-covalently bound light chain beta-2 microglobulin (β2m), and a peptide of eight to ten amino acids. Assembly of HC, β2m, and peptide takes place in the endoplasmic reticulum (ER), followed by transport to the cell surface, where MHC I proteins present the bound peptides to T cell receptors of cytotoxic T cells; they also bind inhibitory receptors on Natural Killer cells. MHC I antigen presentation is central to the cellular immune response of jawed vertebrates against viruses, intracellular parasites, and tumors.

At the cell surface, MHC I proteins exist in three different forms: the HC/β2m/peptide trimers, the ‘peptide-empty’ HC/β2m dimers derived from them by dissociation of the peptide, and the ‘free’ heavy chains derived from the dimers by dissociation of β2m. While the lateral association of MHC I proteins has been observed before (Capps et al., 1993; Arosa et al., 2007; Lu et al., 2012; Matko et al., 1994), other researchers have not detected them (Szöllösi et al., 1989; Damjanovich et al., 1983; Liegler et al., 1991). Optical methods have not provided conclusive evidence which of the three forms are interacting, and biochemical approaches have not clarified where in the cell the binding occurs (Capps et al., 1993; Matko et al., 1994). To add to the complexity, the HC/β2m dimers and the free heavy chains have short half-lives in the cell, which complicates their analysis (Montealegre et al., 2015). For the study of the proposed functions of the MHC I protein-protein interactions (endocytosis, synapse architecture, inflammatory response, receptor modulation; see the discussion), knowledge of location and conformation of the associated proteins is essential (Dixon-Salazar et al., 2014; Chen et al., 2017; Burian et al., 2016; Nizsalóczki et al., 2014; Mocsár et al., 2016).

Our work now shows conclusively that MHC I free heavy chains, but not HC/β2m/peptide trimers or HC/β2m dimers, associate in cis at the plasma membrane.

Results

H-2Kb and H-2Db are specifically captured by antibody micropatterns

Membrane proteins on the surface of living cells can be captured into geometric shapes by antibodies that are printed on to the substrate in micrometer-sized patterns (Figure 1A; [Dirscherl et al., 2017]). We reasoned that any protein that naturally interacts with a captured protein would also be recruited into the patterns, and that this might be used for a protein-protein interaction assay (Schwarzenbacher et al., 2008). We further reasoned that if we printed antibodies that recognize only certain forms of MHC I proteins, the interaction assay might be made specific for certain forms of MHC I proteins.

Specific capture of cell surface Kb on antibody micropatterns.

(A) Schematic presentation of the capture assay. Cells transduced with Kb (red) fused to GFP (green) are incubated on the Y3 antibody micropatterns (anti Kb; magenta). Upon specific antibody-antigen interaction, Kb-GFP is captured on its extracellular epitope by the Y3 antibody pattern elements (see enlargement). (B) Printed antibodies are target-specific. Control experiments demonstrate that Kb-GFP is only captured by the anti-Kb antibody Y3 and not by an antibody specific for Db (27-11-13S). (C) Schematic displaying the different antibody epitopes on the Kb molecule. The Y3 epitope reacts specifically with residues of the α2 helix of Kb-GFP whereas the anti-HA antibody recognizes the additional HA-tag that was N-terminally fused to Kb-GFP. (D) Surface Kb-GFP can be directly captured by the anti-Kb antibody Y3 or by the anti-HA antibody against the N-terminally tagged HA-Kb-GFP. Cells were transduced with Kb-GFP or HA-Kb-GFP and tested for specificity on Y3 or anti-HA antibody micropatterns. Y3 successfully captures both constructs, whereas HA only recognizes the HA-tagged molecules. Scale bar: 25 µm.

https://doi.org/10.7554/eLife.34150.002

We first tested whether the two common β2m-dependent monoclonal antibodies Y3 (which binds to two forms of the murine MHC I allotype H-2Kb, or Kb for short, namely the KbHC/β2m dimers and KbHC/β2m/peptide trimers[Hämmerling et al., 1982]) and 27-11-13S (which binds to two forms of the murine MHC I allotype H-2Db, namely DbHC/β2m dimers and DbHC/β2m/peptide trimers [Ozato and Sachs, 1981]) were still specific for their target allotypes when used in the pattern capture assay. We inked poly(dimethylsiloxane) (PDMS) stamps with solutions of Y3 and 27-11-13S and printed them onto the surface of untreated glass coverslips. We then seeded human STF1 fibroblasts expressing C-terminal green fluorescent protein (GFP) fusions of either Kb or Db onto these coverslips and observed capture of Kb-GFP and Db-GFP by confocal laser scanning microscopy (Figure 1B). As anticipated, Kb-GFP was only captured with Y3, and Db only with 27-11-13S. We conclude that the printed β2m-dependent antibodies still specifically recognize their target allotypes.

In addition to β2m-dependent capture by Y3 or 27-11-13S, we wished to be able to capture MHC I proteins independently of their β2m or peptide association. Thus, we next tested whether MHC I proteins can also be captured via an N-terminal (extracellular) influenza hemagglutinin (HA) epitope tag (Figure 1C, bottom). We printed patterns of the monoclonal anti-HA antibody 12CA5 and seeded STF1 cells expressing either a HA-Kb-GFP fusion construct or Kb-GFP, which lacked the HA epitope. As expected, only HA-Kb-GFP was captured, but not Kb-GFP (Figure 1D). The HA tag did not interfere with the capture of HA-Kb-GFP on Y3 antibody micropatterns (Figure 1D). We conclude that the anti-HA antibody can be used to specifically capture HA-tagged MHC I proteins.

Stabilizing effect of conformation-specific antibodies allows for differential patterning of Kb dimers and free heavy chains

We next sought to establish conditions in which KbHC/β2m dimers, without peptide, are preferentially captured in the antibody patterns. In STF1 cells, which lack TAP (the transporter associated with antigen processing) and thus cannot load MHC I proteins with high-affinity peptides in the ER (de la Salle et al., 1999), such dimers can be accumulated at the cell surface by incubation at 25°C (Ljunggren et al., 1990; Montealegre et al., 2015). These peptide-receptive KbHC/β2m dimers can be detected by subsequent binding of fluorescently labeled peptide (Saini et al., 2013).

We therefore printed Y3 on glass coverslips, seeded STF1 cells expressing HA-Kb-GFP onto the patterns, and then incubated at 25°C overnight. Next morning, we added the Kb-specific peptide SIINFEKL(abbreviated SL8) labeled with the TAMRA fluorophore (SL8TAMRA). We observed a striking patterned staining of the fluorescent peptide, which demonstrates that peptide-receptive Kb2m dimers had been captured in the patterns (Figure 2A, column 2). To show that binding of the peptide was specific, we pre-incubated the cells with unlabeled SIINFEKL peptide, which blocked SL8TAMRA binding (Figure 2A, column 3). Thus, we were able to capture KbHC/β2m dimers into patterns and subsequently bind peptide to them.

Antibody micropatterns determine stability of the captured Kb population.

(A) Cells expressing HA-Kb-GFP were captured on Y3 or anti-HA antibody micropatterns and incubated at 25 or 37°C to allow for the dissociation of β2m. To identify the nature of the captured Kb-GFP population (green channel), specific fluorescent peptide SIINFEKL (SL8TAMRA; red channel) was added to the samples. Based on their ability to bind peptide, one can distinguish between the peptide-receptive KbHC/β2m dimer and the Kb free heavy chains, which are incapable to bind peptide. (B) For further characterization of the captured HA-Kb-GFP on Y3 or anti-HA antibody micropatterns, immunostaining experiments were performed. Immunostaining of captured HA-Kb-GFP molecules with the anti-β2m antibody (BBM.1Atto542) reveals dissociation of β2m from anti-HA antibody micropatterns at 37°C (column 3). Addition of the specific ligand peptide SIINFEKL (SL8) during 37°C incubation prevents β2m dissociation (column 4). Scale bars: 25 µm.

https://doi.org/10.7554/eLife.34150.003

We then repeated the same experiment on patterns of anti-HA antibody, with the same result (Figure 2A, column 5). Thus, both Y3 and anti-HA antibodies captured peptide-receptive KbHC/β2m dimers on the surface of the STF1 cells.

We next wished to dissociate β2m from the captured KbHC/β2m dimers in order to obtain patterned free heavy chains. At 37°C, dissociation of β2m from KbHC/β2m dimers occurs rapidly, whereas at 25°C, the rate of dissociation of β2m is significantly reduced (Montealegre et al., 2015; Day et al., 1995). Thus, we repeated the above experiments on anti-HA and Y3 patterns, but before adding the SL8TAMRA peptide, we shifted the cells to 37°C for two to three hours. As expected, after incubation at 37°C, HA-Kb-GFP patterns showed no binding of SL8TAMRA (Figure 2A, column 4), which shows that the HA-captured MHC I proteins lost their peptide-binding capacity, probably due to the dissociation of β2m. Very interestingly, HA-Kb-GFP captured with Y3 retained its ability to bind peptide at 37°C (Figure 2A, column 1). This suggests that the β2m-dependent Y3 antibody stabilizes the KbHC/β2m dimer complex to which it is bound by preventing the dissociation of β2m, similar to the action of peptide (Townsend and Bodmer 1989).

To test the hypothesis that the lack of peptide binding of the Kb proteins that were captured by HA and incubated at 37°C was due to the loss of β2m, we repeated the same experiment, but instead of adding fluorescent peptide, we fixed the cells and immuno-stained with the anti-β2m antibody BBM.1 that was directly labeled with Atto 542 (BBM.1Atto542) (Figure 2B). As predicted, the Y3 micropatterns stained positive for β2m at both temperatures (Figure 2B, columns 1 and 2), whereas anti-HA micropatterns stained for β2m only at 25°C (Figure 2B, columns 3 and 5). When we added SIINFEKL to the cells on anti-HA-micropatterns before shifting them to 37°C, we observed that BBM.1Atto542 staining was restored in these samples (Figure 2B, column 4). We conclude that captured Kb free heavy chains can be generated by inducing the dissociation of β2m from KbHC/β2m dimers captured on anti-HA patterns.

Taken together, we are able to selectively hold three different forms of Kb on the surface of STF1 cells: free Kb heavy chains (at 37°C on HA patterns), KbHC/β2m dimers (at 25°C on anti-HA patterns, or at 25 or 37°C on Y3 patterns), and KbHC/β2m/peptide trimers (by addition of peptide on anti-HA and Y3 patterns).

Antibody micropatterns reveal conformation-dependent in cis interactions of Kb free heavy chains

Since we were able to distinguish the three different forms of Kb held in the patterns, we next investigated whether any of these forms associate in cis on the plasma membrane. For this, we designed a two-hybrid assay (Figure 3A): One Kb construct had an N-terminal HA tag but no GFP (HA-Kb), the other carried a C-terminal GFP but no HA tag (Kb-GFP). We reasoned that HA-Kb would be captured by the anti-HA antibodies, but the GFP pattern would only become detectable by microscopy if HA-Kb and Kb-GFP interacted together, since Kb-GFP alone is not captured by anti-HA micropatterns (Figure 1D).

Figure 3 with 2 supplements see all
Antibody micropatterns reveal conformation-dependent in cis interaction of captured Kb-GFP.

(A) For the two-hybrid assay, cells were co-transduced with two Kb constructs: Kb with an N terminal HA tag (HA-Kb) and Kb-GFP (GFP fused to the cytoplasmic tail). (B) Cells were incubated on anti-HA or Y3 antibody micropatterns at different temperatures. Recruitment of Kb-GFP (green channel) to the anti-HA antibody micropatterns occurs specifically at 37°C and can be inhibited by addition of the SIINFEKL (SL8) peptide (column 3 and 4). The single chain mutant, scKb-GFP (which has β2m covalently linked to the Kb heavy chain) is also not recruited to the antibody micropatterns (column 5). From top to bottom: Antibody micropatterns, Kb-GFP, phase contrast, and schematic representation. Scale bar: 25 µm. (C) For quantification of co-capture, the mean fluorescence intensities of Kb-GFP of the total cell and the areas of pattern elements were determined. The redistribution of Kb-GFP leads to increased fluorescence intensity levels in the areas of the pattern elements and is represented as an increase of the ratio of the fluorescence intensity of the pattern elements over the fluorescence intensity of the entire cell (see Materials and methods). The plot shows the mean (red) ± SEM and the distribution of the calculated ratios from individual cells (black symbols; n (cells) ≥ 14) of ≥ 2 independent experiments (****: Significant difference, p<0.0001, two-tailed t-test).

https://doi.org/10.7554/eLife.34150.004

To perform the experiment, we co-transduced STF1 cells with HA-Kb and Kb-GFP, seeded the cells on anti-HA micropatterns, and incubated them overnight at 25°C to accumulate KbHC/β2m dimers of both Kb constructs at the cell surface. The next day, we either left them at 25°C or shifted them to 37°C and followed the patterning of the GFP fluorescence.

We observed strong co-patterning of both forms after 37°C incubation, suggesting that free heavy chains can interact in cis in the membrane (Figure 3B, column 3). When we inhibited dissociation of β2m by addition of SIINFEKL peptide (Figure 3B, column 4) or incubation at 25°C (Figure 3B, column 6), co-patterning was abolished. (As a control, on Y3 patterns, in contrast, where both HA-Kb and Kb-GFP are directly bound by the antibody, strong patterning of Kb-GFP was visible even in the presence of SIINFEKL (Figure 3B, columns 1 and 2)). These data show that Kb does not associate in cis as long as β2m is bound.

To further test this conclusion, we co-transfected STF1 cells with HA-Kb and scKb-GFP, a single-chain construct in which the Kb heavy chain and β2m are linked by a peptide linker such that β2m cannot dissociate (Montealegre et al., 2015). As in the previous control experiment, no co-patterning was observed (Figure 3B, column 5). These controls also demonstrated that the in cis interaction of the Kb free heavy chains was not simply induced by the GFP domain of the Kb-GFP fusion proteins.

To demonstrate that the non-fluorescent HA-Kb molecules were indeed present in the patterns together with Kb-GFP, we repeated the experiment in Figure 3B (column 3) with the construct E3-HA-Kb, in which an additional tag of a 21 amino acids (the E3 tag; see Materials and methods) is attached to the N terminus of HA-Kb. This tag specifically binds to the fluorescently labeled synthetic peptide, K4Atto633. After co-patterning of Kb-GFP was observed, we additionally stained with K4Atto633 and found close colocalization with Kb-GFP in the patterns (Figure 3—figure supplement 1). We conclude that the free heavy chain of E3-HA-Kb is indeed captured on the patterns and then recruits the free heavy chain of Kb-GFP by an in cis interaction. We quantified this recruitment by measuring both the mean fluorescence intensity (MFI) of the entire cell and the MFI of the pattern elements in the micrographs of Figure 3B to standardize the increase in fluorescence signal on the pattern elements upon redistribution of Kb-GFP by its in cis interaction with HA-Kb (see Materials and methods and Figure 3C).

Taken together, our data demonstrate that Kb free heavy chains, but not KbHC/β2m dimers or KbHC/β2m/peptide trimers, associate in cis in the plasma membrane of live cells.

In murine (and human) cells, up to six MHC I allotypes co-exist at the cell surface. We thus wished to investigate heterotypic in cis interactions between free heavy chains of Kb and Db. In an experiment with HA-Kb and Db-GFP performed in analogy to Figure 3B above, we observed weaker but still specific (initiated by the 37°C shift) recruitment of Db-GFP to the antibody pattern elements, indicating heterotypic in cis interactions between the two allotypes Kb and Db (data not shown).

In cis interactions of Kb confirmed by co-immunoprecipitation

We next tested this finding in a co-immunoprecipitation experiment, without the use of micropatterns. We used the same STF1 cells co-transfected with HA-Kb and Kb-GFP and incubated them at 25°C overnight to accumulate both Kb constructs at the cell surface. The cells were then shifted to 37°C for 15 min (the half-life of KbHC/β2m dimers at the cell surface [Montealegre et al., 2015]) to dissociate β2m and initiate co-localization. Then, the cells were trypsinized and lysed, and HA-Kb was immunoprecipitated with the anti-HA antibody. We found efficient co-precipitation of Kb-GFP with HA-Kb, which was abolished if SIINFEKL peptide was added to the cells during the 37°C incubation (Figure 4). Thus, just like in the micropattern assay, the peptide clearly inhibited the interaction, suggesting that only free heavy chains were co-precipitating.

Figure 4 with 1 supplement see all
In cis interactions of Kb-GFP and HA-Kb are peptide-dependent and generally occur in cells at 37°C.

For co-immunoprecipitation, the same co-transduced cells from the previous experiment were used (STF1/HA-Kb and Kb-GFP). Cells were incubated at 25°C overnight to increase Kb cell surface levels and then shifted to 37°C to allow for the dissociation of β2m from the Kb heavy chain. The SIINFEKL (SL8) peptide was added as control to inhibit β2m dissociation (lanes 2 and 4). Cells were then lysed and successfully immunoprecipitated with an anti-HA antibody (bottom row). Immunoisolates and lysate control samples were analysed by western blotting by sequential staining with an anti-GFP antibody (top row) and an anti-HA antibody (bottom row). The Kb-GFP construct was specifically co-immunoprecipitated in the absence of peptide, similar to the result on antibody micropatterns (lane 1).

https://doi.org/10.7554/eLife.34150.008

Discussion

We have developed antibody micropatterns into a novel two-hybrid assay for the detailed investigation of conformation-specific in cis interactions of our model protein, MHC I. The versatility of our assay with extracellular HA and intracellular GFP fusions supports a broad range of applications, especially in the study of specific protein-protein interactions that require investigation in the native environment of live cells. This two-hybrid assay circumvents the disadvantages of employing two different fluorescent proteins on the cytoplasmic tails of the proteins of interest (for example for FRET microscopy) that may be biased through unspecific interactions and aggregation of the fluorescent tags.

The example of MHC I proteins demonstrates the challenges of the functional analysis of protein-protein interactions and the limitations of conventional methods, which yield no information on the spatial resolution or the distinction of different protein conformations. Previous experiments with FRET have revealed cluster formation of antibody-labelled MHC I proteins at the cells surface, but the involvement of free heavy chains was only indirectly shown (Matko et al., 1994). Other studies involved co-immunoprecipitation experiments that revealed the existence of free heavy chain-dimers of different murine MHC I allotypes by pull-down with conformation-specific antibodies. However, it could not be excluded that the detected interactions were enhanced, or indeed caused, by detergents after cell lysis. Additional pulse-chase experiments confirmed that MHC I proteins associate after they have traversed the medial Golgi, but could not localize them to the cell surface (Capps et al., 1993).

Our own co-immunoprecipitation experiments confirm these observations for the murine Kb allotype, but they also lack spatial resolution (Figure 4). The differential co-immunoprecipitation of surface-biotinylated MHC I proteins finally confirms the presence of protein-protein associations at the cell surface (Figure 4—figure supplement 1), but even this method cannot entirely exclude the involvement of intracellular MHC I proteins.

Our two-hybrid assay finally solves the questions of generation and location for MHC I protein-protein interactions. The assay principle has allowed us to establish a system in which we generate defined conformations of MHC I proteins. The results demonstrate that MHC I protein association depends on the generation of free heavy chains, and together with our previous work (Montealegre et al., 2015), they suggest that these free heavy chains originate from the captured empty KbHC/β2m dimers at the cell surface upon dissociation of β2m. By holding the dimers in the anti-HA patterns and triggering the dissociation of β2m with the 37°C shift, we were able to show that in cis interactions are indeed happening at the cell surface (Figure 3B). Of course, it is possible that in addition, free heavy chains are generated elsewhere in the cell by β2m dissociation, for example in endosomes, and that they might associate in these locations also.

Is it possible that more than one kind of MHC I protein-protein interaction exists in cells? The MHC I clusters identified by the Yewdell and Edidin groups (Yewdell, 2006; Matko et al., 1994) contain peptide-bound MHC I proteins. Since our cells were TAP-deficient and thus contained no, or few, peptides for binding to MHC I proteins, we would not have seen the clusters observed by them. If, for example, the Yewdell clusters are formed in the ER then they would not even have formed in our system upon addition of external peptide.

The spatial organization of bait proteins in the plasma membrane allows for the quantification of co-captured proteins into the antibody pattern elements (Figure 3C). In our setup, we determined the distribution of prey proteins in control experiments as biological background (this corresponds to our background ratio of 1.1). This background includes also MHC I proteins that are co-captured on the antibody pattern elements before the temperature shift. Whether this background corresponds to pre-formed protein-protein interactions or a heterogeneous surface distribution was not tested and requires detailed analysis.

For MHC I in cis interactions at the cell surface, various functional roles have been proposed. They might be a means of accelerated disposal for free heavy chains, preventing re-binding of β2m and peptide and perhaps leading to enhanced internalization and degradation in lysosomes (Montealegre et al., 2015). This hypothesis is supported by our finding that associated MHC I proteins do not bind peptide well (Figure 3—figure supplement 2) and therefore probably do not interact with TCRs. They might be in trans ligands for NK cell receptors or similar proteins, perhaps signaling stress or activation states (Garcia-Beltran et al., 2016; Burian et al., 2016). We cannot entirely exclude that the associated MHC I proteins contain some KbHC/β2m dimers that are peptide-receptive, as has been suggested for human MHC I oligomers (Bodnár et al., 2003), but free heavy chains are clearly essential for in cis interactions, since single-chain KbHC/β2m dimers do not associate (Figure 3A). Another possibility is that the associated free heavy chain might influence the surface levels of other proteins with which MHC I proteins are known to interact, such as APLP or insulin receptor, thereby mediating non-immunological functions of MHC I proteins (Tuli et al., 2008; Shatz, 2009; Dixon-Salazar et al., 2014). Our assay is a promising tool to extend the interaction studies for MHC I proteins by the proposed interaction partners.

We have shown here the formation of homotypic in cis interactions of the murine MHC I allotype Kb. Interestingly, previous work suggests that the tendency of in cis interactions varies among MHC I allotypes (Capps et al., 1993). Thus, it was hypothesized that those MHC I allotypes that do not associate are not internalized (by the accelerated disposal mechanism proposed above) and that they will bind exogenous peptides to provoke autoimmune reactions (Capps et al., 1993). This might be interesting in the case of those subtypes of HLA-B27 that are implicated in inflammatory autoimmune diseases such as spondyloarthropathies (Chen et al., 2017; Allen et al., 1999). For HLA-B*27:05, formation of heavy chain dimers at the cell surface, or in early endocytic compartments, was demonstrated to occur through a disulfide bond between the unpaired cysteine-67 residues. Since Kb does not have an unpaired cysteine in the extracellular domain, this type of dimerization is not possible for Kb. Still, the interesting possibility exists that the in cis heavy chain associations described by us for Kb might also occur with B*27:05 and might cause the formation of the covalent B*27:05 dimers. We look forward to future investigations.

Due its versatility, our assay allows for the development of a screen to test for the tendency of individual MHC I allotypes to associate with themselves, and with other allotypes. This may be extended to human MHC I proteins, whose empty dimers can be enriched at the cell surface by incubation with low-affinity dipeptide ligands (Saini et al., 2015), and even to the empty forms of HLA-F that were recently discovered to bind NK cell receptors (Garcia-Beltran et al., 2016; Burian et al., 2016). Consequently, by its application to the human system, this screening tool can be developed to investigate the correlation between cell surface protein-protein interactions and human autoimmune disease. Generation of anti-HA antibody micropatterns by microcontact printing on conventional glass coverslips makes this assay especially suitable for such high-throughput approaches.

In addition to its demonstrated application in the detection of conformation-dependent in cis interactions, the assay can be further developed towards more detailed analysis. One possibility is the integration of conventional immunostaining for the identification of other proteins involved in the redistribution of co-captured proteins. For MHC I proteins, for example, observing the accumulation of adaptor proteins involved in endocytic processes (e.g. Rab proteins) on the pattern elements under condition of co-capture will contribute to understand the nature of MHCI protein endocytosis and the functional role of in cis interactions.

Another possible technical development is to combine the assay with fluorescence revovery after photobleaching (FRAP) measurements to test the dynamics of the interactions (i.e. dissociation and re-association). Such a combined two-hybrid-FRAP assay is potentially superior to conventional (FRET) experiments, since the enrichment of proteins in the pattern elements increases the abundance of the interaction partners and might thus enable the detection of very weak interactions.

Materials and methods

Key resources table
Reagent type (species)
or resource
DesignationSource or referenceIdentifiersAdditional information
Gene
(mus musculus)
H-2KbNCBI GenBankNM_001001892.2
Gene
(mus musculus)
H-2DbNCBI GenBankNM_010380.3
Cell line (homo sapiens)STF1PMID:10074495N/A
Transfected construct
(mus musculus)
STF1/Kb-GFPThis paperN/AStable cell lines were generated
by lentiviral transduction and
antibiotic selection as described
in PMID: 24806963.
Transfected construct
(mus musculus)
STF1/Db-GFPThis paperN/A
Transfected construct
(mus musculus)
STF1/E3 HA-Kb-GFPThis paperN/A
Transfected construct
(mus musculus)
STF1/E3 HA-Kb
+Kb-GFP
This paperN/A
Transfected construct
(mus musculus)
STF1/E3 HA-Kb
+Kb-hβ2m-GFP
(single chain)
This paperN/A
AntibodyY3PMID: 6181513N/AProduced and purified in house from
hybridoma cells. 0.6 µg/µL for printing
Antibody27-11-13SPMID: 6935293N/A
Antibody12CA5
(Hemagglutinin; HA)
PMID: 6192445N/AProduced and purified in house from
hybridoma cells 0.6 µg/µL for printing;
1:100 dilution of hybridoma supernatant
for Western blotting.
AntibodyBBM.1PMID: 91522N/AProduced and purified in house from
hybridoma. 0.1 µg/µL for staining.
Antibodyrabbit anti-GFPAbcamCat #: ab290,
RRID:AB_303395
1:1000
Antibodyrabbit antisera against
H-2Kand H-2Db
Charles River
Laboratories
Rabbits were immunized
with a peptide
corresponding to
residues 331–349 of
the cytoplasmic tail
of both heavy chains
1:1000
Antibodygoat anti-rabbit
IgG-AP
Bio-RadCat #: 170–6518,
RRID:AB_11125338
1:10000
Antibodydonkey anti-mouse
IgG Alexa Fluor 568
Thermo Fisher
Scientific
Cat #: A10037,
RRID:AB_2534013
1:400 for staining antibody
micropatterns
Recombinant DNA reagentpuc2CL6Ipwo
(lentiviral vector)
PMID: 21248040N/A
Recombinant DNA reagentpuc2CL6IPwo/
Kb-GFP (plasmid)
This paperN/AC-terminally GFP-tagged H-2Kb and
H-2Db cDNA were cloned into the
lentiviral vector via the XhoI and AgeI sites.
Recombinant DNA reagentpuc2CL6IPwo/
Db-GFP (plasmid)
This paperN/A
Recombinant DNA reagentpuc2CL6IPwo/E3
-HA-Kb-GFP (plasmid)
This paperN/AN-terminally E3-tagged and HA-tagged)
H-2Kb (±GFP) cDNA were cloned into the
lentiviral vector via the XhoI and AgeI sites.
Recombinant DNA reagentpuc2CL6IPwo/
E3-HA-Kb (plasmid)
This paperN/A
Recombinant DNA reagentpuc2CL6IPwo/
E3-HA-Kb (plasmid)
This paperN/A
Recombinant DNA reagentpuc2CL6IPwo/Kb-
2m-GFP (plasmid)
This paperN/ACloning strategy of the single chain
construct according to PMID: 16049493
Sequence-based reagentforward primer E3 tagThis paperN/A5´-ACTCAGACCCGCGCGGGCGAGATCG
CAGCTCTGGAGAAGGAGATTGCCGCAT
TGGAGAAGGAGATAGCGGCACTCGAG
AAGTATCCATACGACGTCCC-3´
Sequence-based reagentreverse primer
E3 tag
This paperN/A5`-GGGACGTCGTATGGATACTTCTCGA
GTGCCGCTATCTCCTTCTCCAATGCGG
CAATCTCCTTCTCCAGAGCTGCGAT
CTCGCCCGCGCGGGTCTGAGT-3´
Sequence-based reagentforward primer HA
tag (1/2)
This paperN/A5`-CCGACTCAGACCCGCGCGGGCC
CATATCCATACGACGTCCCACACTC
GCTGAGGTATTTCGTCACC-3´
Sequence-based reagentforward primer HA
tag (1/2)
This paperN/A5`-GGCCCATATCCATACGACGTCCCAG
ATTATGCCGGCGGTGGACACTCGCTG
AGGTATTTCGTCACC-3´
Sequence-based reagentreverse primer HA tagThis paperN/A5`-GGTGACGAAATACCTCAGCGAGTG
TGGGACGTCGTATGGATATGGGCCCG
CGCGGGTCTGAGTCGG-3´
Sequence-based reagentreverse primer HA tagThis paperN/A5`-GGTGACGAAATACCTCAGCGAGTG
TCCACCGCCGGCATAATCTGGGACGTCG
TATGGATATGGGCC-3´
Peptide, recombinant proteinSIINFEKL peptideGeneCust Ellange,
Luxemburg
N/A2 µM final concentration
Peptide, recombinant proteinSIINFEKLTAMRA peptideGeneCust Ellange,
Luxemburg
N/A2 µM final concentration
Peptide, recombinant proteinK4Atto 633 peptideemc microcollectionsN/A25 nM final concentration
Peptide, recombinant proteinK4bio peptideemc microcollectionsN/A200 nM final concentration
Chemical compound, drugAlexa Fluor
647 NHS ester
Thermo Fisher ScientificCat #: A37566
Chemical compound, drugAtto 542
NHS ester
ATTO-TEC,Cat #: AD 542–31
Software, algorithmImageJNational Institutes
of Health

Photolithography

Silicon master molds were prepared by semiconductor photolithography as described previously. See Dirscherl et al. (2017) for details.

PDMS stamps and antibody patterns

PDMS stamps were generated from basic elastomer and curing agent (Sylgard 184 Silicone Elastomer Kit) from Dow Corning (Midland, USA) mixed in a 10:1 ratio. The prepared stamps were inked with the indicated antibody solutions and then placed on round microscopy glass coverslips (#1, 22 mm). See Dirscherl et al. (2017) for details.

Patterning cell surface proteins

Coverslips were placed into sterile 6-well plates. Cells were immediately seeded as indicated at a concentration of ≈50.000 cells per well and incubated on the antibody patterns. Usually, cells were incubated for 4–6 hr at 37°C for adhesion and then shifted to 25°C overnight to accumulate Kb molecules at the cell surface for a better signal-to-noise ratio of patterned Kb molecules. For co-capturing experiments, samples were then kept at 25° C or shifted back to 37°C for 3–4 hr to allow for the dissociation of β2m.

For each experiment fresh antibody micropatterns are generated, representing technical replicates. The experiments were individually repeated on different days at least three times and representative images are shown in the respective figures. For each experimental condition, at least triplicates were performed per experiment. In this setup, the individual cells seeded onto the antibody micropatterns represent the biological replicates.

Antibodies

Mouse monoclonal hybridoma supernatants Y3 (against the complex of Kb free heavy chain with β2m (Hämmerling et al., 1982), 27-11-13S (against the complex of Db free heavy chain with β2m (Ozato and Sachs, 1981), hemagglutinin (HA) 12CA5 (Niman et al., 1983), and BBM.1 (Brodsky et al., 1979) were described previously. Antibodies for immunoprecipitation were rabbit anti-GFP (Abcam ab290), rabbit antisera against H-2Kb and H-2Db (Charles River Laboratories, Kisslegg, Germany), and goat anti-rabbit IgG-AP conjugate (1706518, Biorad, Munich, Germany). Secondary antibody against the HA-antibody was donkey anti-mouse IgG Alexa Fluor 568 (a10037, Thermo Fisher Scientific, Darmstadt, Germany).

Dyes

Purified antibodies were either labeled with the Alexa Fluor−647 NHS ester (Y3, 27-11-13S and 12CA5) or with the Atto 542 NHS ester (BBM.1) according to the manufacturers´ protocols. Alexa Fluor−647 NHS was obtained from Thermo Fisher Scientific (Darmstadt, Germany) and the Atto 542 NHS from ATTO-TEC (Siegen, Germany).

Peptides

Peptides were synthesized by GeneCust (Ellange, Luxemburg) and emc microcollections (Tübingen, Germany) and purified by HPLC (90% purity). The Kb-specific peptide SL8 (SIINFEKL in the single-letter amino acid code) was labeled with the fluorescent dye 5’-carboxytetramethylrhodamine (TAMRA) on the lysine side chain (avoiding interference with peptide binding to Kb) to give SIINFEKLTAMRA. Labeled and unlabeled peptides were added to the cells at a final concentration of 2 µM for 15–30 min at 37°C to detect peptide binding. Cells were then washed with phosphate buffered saline (PBS, 10 mM phosphate pH 7.5, 150 mM NaCl), fixed, and observed by confocal laser scanning microscopy (cLSM).

Cell lines and gene expression

For experiments, we used STF1 TAP-deficient human fibroblasts (kindly provided by Henri de la Salle, Etablissement de Transfusion Sanguine de Strasbourg, Strasbourg, France; see (de la Salle et al., 1999) for reference). These cells were chosen to make sure that only the murine allotype H-2Kb was investigated, without interference of other murine MHC I allotypes that might cross-react with the used antibodies. STF1 cells were authenticated by HLA haplotype genotyping; they were regularly tested for mycoplasma. The cells were stably transduced with Kb-GFP, Db-GFP, and E3-HA-Kb-GFP. Lentiviruses were produced and used for gene delivery as described previously (Hein et al., 2014). For co-capturing experiments, co-transduced STF1 cells were used. The cells were first selected with puromycin for E3-HA-Kb (the additional E3 tag (EIAALEK)3 is a 21 amino-acid long extracellular tag that was initially introduced for co-staining experiments). Cells were then transduced with the indicated Kb-GFP constructs and hβ2m where indicated and again selected with puromycin to obtain STF1/E3-HA-Kb+Kb GFP or STF1/E3-HA-Kb+Kb-hβ2m-GFP (single chain Kb, single chain construct in which the light chain β2m is fused by a linker to the Kb heavy chain).

Staining with the K4 peptide

The K4 peptide was synthesized by emc microcollections (Tübingen, Germany). The E3-tag specific peptide K4 (Litowski and Hodges, 2002; Yano et al., 2008) (a 28 amino-acid long peptide abbreviated as (KIAALKE)4 in the single-letter amino acid code) was labeled with an Atto 633 fluorophore at the N terminus. For co-staining experiments, cells were fixed with 3% paraformaldehyde (PFA), washed, and permeabilized with 0.1% Triton X-100. The K4Atto633 peptide was added to the cells at a final concentration of 25 nM in PBS and incubated for 5 min at RT to stain the transduced E3-HA-Kb construct.

Immunofluorescence stainings

For antibody co-staining experiments, cells were fixed with 3% PFA, washed and permeabilized with 0.1% Triton X-100 and stained with the respective antibodies.

Microscopy

A confocal laser scanning microscope (LSM 510 Meta, Carl Zeiss Jena GmbH, Germany) equipped with argon and helium-neon lasers at 488, 543 and 633 nm was used. Images were recorded with a 63x Plan Apochromat oil objective (numerical aperture 1.4) at a resolution of 1596 × 1596 pixels. Data acquisition was performed with the LSM 510 META software, release 3.2 (Carl Zeiss Jena GmbH). During image acquisition, patterns and cells were imaged in the same focal plane at a pinhole of ≈1 Airy unit. Image analysis and processing were performed using ImageJ (National Institutes of Health, Bethesda, USA). Image processing comprises cropping, rotation and adjustment of brightness and contrast levels.

Generally, all cells on the antibody micropatterns were scanned by eye, and representative cells were chosen for image acquisition. Due to the range of expression levels, only cells with a moderate expression level were selected for evaluation. Only adhered cells were evaluated; cells with altered morphologies (e.g. apoptotic cells) were excluded.

Quantification of co-captured proteins

To compare the spatial distribution of Kb-GFP between the pattern elements and the pattern element interspaces, we determined the mean fluorescence intensity of GFP in the entire cell area and also for the areas on the pattern elements (ImageJ, National Institutes of Health, Bethesda, USA). The ratio of fluorescence intensity on pattern elements over total fluorescence intensity was calculated. Control experiments (peptide addition, incubation at 25°C and the single chain construct) gave a 1.1 ratio, and thus this was defined as the background signal. The co-capturing experiments usually showed a ratio of 1.3 (Figure 3C). For each condition, ≥ 14 cells of ≥ 2 individual experiments were used.

Co-immunoprecipitation

For co-immunoprecipitation with the anti-HA antibody, co-transduced (E3-HA-Kb+Kb-GFP) and selected STF1 cells were incubated at 25°C overnight. The next day, cells were incubated in presence or absence of 10 µM SL8 for 10 min at 25°C, then shifted to 37°C for 15 min, trypsinised and harvested. Cell pellets were lysed in native lysis buffer (50 mM Tris-Cl (pH 7.4), 150 mM NaCl, 5 mM EDTA, and 1% Triton X-100) for 1 hr at 4°C. After lysis, the supernatant was immunoprecipitated with the anti-HA antibody for 30 min at 4°C. Beads were washed and resuspended in Laemmli sample buffer (LSB) buffer and boiled at 95°C for 10 min. The immunoisolates were separated by SDS-PAGE and immunoblotted sequentially with an anti-GFP antibody and an anti-HA antibody. The experiment was performed three times.

For co-immunoprecipitation of the cell surface Kb molecules with the biotinylated K4 peptide (K4biotin), co-transduced (E3-HA-Kb+Kb-GFP) and selected STF1 cells were incubated at 25°C overnight. Next day, cells were incubated in presence or absence of 10 µM SL8 for 10 min at 25°C. E3-tagged Kb-molecules were labeled with K4biotin for 5 min at room temperature using 200 nM biotinylated K4 peptide. Following biotinylation, cells were placed at 37°C for 15 min to allow for co-capture. Cells were collected into native lysis buffer (50 mM Tris-Cl (pH 7.4), 150 mM NaCl, 5 mM EDTA, and 1% Triton X-100) by scraping and lysed for 45 min at 4°C. Biotinylated E3-HA-Kb (with K4biotin) was immunoprecipitated from post-nuclear supernatants using neutravidin-agarose beads (Thermo Fisher Scientific, Darmstadt Germany). Beads were washed in lysis buffer and wash buffer (50 mM Tris-Cl (pH 7.4), 150 mM NaCl, 5 mM EDTA, and 0.1% Triton X-100) and supplemented with endogylcosidase F1 for 2 hr at 37°C, or left untreated. Isolated proteins were retrieved from beads by boiling and separated by SDS-PAGE. Kb-GFP was detected by anti-GFP antiserum and E3-HA-Kb was detected by 12CA5 (anti-HA) following western blotting. The experiment was performed twice.

References

  1. 1
    Cutting edge: HLA-B27 can form a novel beta 2-microglobulin-free heavy chain homodimer structure
    1. RL Allen
    2. CA O'Callaghan
    3. AJ McMichael
    4. P Bowness
    (1999)
    Journal of Immunology 162:5045–5048.
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
    In vivo dimeric association of class I MHC heavy chains. Possible relationship to class I MHC heavy chain-beta 2-microglobulin dissociation
    1. GG Capps
    2. BE Robinson
    3. KD Lewis
    4. MC Zúñiga
    (1993)
    Journal of Immunology 151:159–169.
  7. 7
    Molecular Medicine Reports, 15
    1. B Chen
    2. J Li
    3. C He
    4. D Li
    5. W Tong
    6. Y Zou
    7. W Xu
    (2017)
    D.A. Spandidos.
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
    Protein micropatterns printed on glass - novel tools for protein-ligand binding assays in live cells
    1. C Dirscherl
    2. S Springer
    (2017)
    Engineering in Life Sciences, 18, 10.1002/elsc.201700010.
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
    Clustering of class I HLA molecules on the surfaces of activated and transformed human cells
    1. J Matko
    2. Y Bushkin
    3. T Wei
    4. M Edidin
    (1994)
    Journal of Immunology 152:3353-60.
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
    Monoclonal antibodies to mouse MHC antigens. III. Hybridoma antibodies reacting to antigens of the H-2b haplotype reveal genetic control of isotype expression
    1. K Ozato
    2. DH Sachs
    (1981)
    Journal of Immunology 126:317-21.
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
    Physical association between MHC class I and class II molecules detected on the cell surface by flow cytometric energy transfer
    1. J Szöllösi
    2. S Damjanovich
    3. M Balàzs
    4. P Nagy
    5. L Trón
    6. MJ Fulwyler
    7. FM Brodsky
    (1989)
    Journal of Immunology 143:208–213.
  33. 33
  34. 34
    Methods in Enzymology
    1. J Weghuber
    2. M Brameshuber
    3. S Sunzenauer
    4. M Lehner
    5. C Paar
    6. T Haselgrübler
    7. M Schwarzenbacher
    8. M Kaltenbrunner
    9. C Hesch
    10. W Paster
    11. B Heise
    12. A Sonnleitner
    13. H Stockinger
    14. GJ Schütz
    (2010)
    133–151, Detection of Protein–Protein Interactions in the Live Cell Plasma Membrane by Quantifying Prey Redistribution upon Bait Micropatterning, Methods in Enzymology, 10.1016/S0076-6879(10)72012-7.
  35. 35
  36. 36

Decision letter

  1. Michael L Dustin
    Reviewing Editor; University of Oxford, United Kingdom
  2. Arup K Chakraborty
    Senior Editor; Massachusetts Institute of Technology, United States

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "A novel two-hybrid antibody micropattern assay reveals conformation-specific cell surface clustering of MHC I proteins" for consideration by eLife. Your article has been reviewed by three peer reviewers, including Michael L Dustin as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Arup Chakraborty as the Senior Editor.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

The reviewers agreed that the use of micro contact printing to evaluate lateral interactions MHC class I proteins in situ was novel and presented a useful new approach to such problems. There was a consensus that the tool would be more powerful if the patterns were quantified and it was felt that this would be straight forward as the patterns are user defined and should be easy to segment and measure. This might allow application to weaker interactions where results may not as obvious.

Essential revisions:

1) Quantify imaging results and validate for this example. Given this signal to noise, how weak an association could be detected?

2) We would strongly recommend not using the word cluster to describe the patterning of class I molecules with glass bound Abs. While this is clearly a useful technique, it is highly artificial, and for all we know, each "cluster" could be a single pair of co-localized Abs. Cluster has been used (appropriately) in the past to refer to spontaneously concentrated class I molecules, likely present in membrane domains that limit their ability to diffuse from each other. The use of cluster in the present study muddies the waters considerably, and will likely increase confusion, rather than the opposite.

3) Is it possible to reverse the association of non-conformed Kb molecules by addition peptide and or β2m?

4) Subsection “Stabilizing effect of conformation-specific antibodies allows for differential patterning of Kb dimers and free heavy chains”, first paragraph: regarding the use of TAP negative cells (whose original publication should be cited), Day et al., 1994, reported that TAP positive cells actually express more peptide receptive cell surface molecules, undermining the reasoning of these statements.

5) "So their endogenous class I molecules would not interfere with the Ab micropatterns". This should be rephrased as it is not clear.

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Thank you for submitting your article "A novel two-hybrid antibody micropattern assay reveals conformation-specific cell surface clustering of MHC I proteins" for consideration by eLife. Your article has been reviewed by two peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Arup Chakraborty as the Senior Editor. The reviewers have opted to remain anonymous.

Summary:

The original reviewers agree that you have provided a thoughtful response to essential revisions proposed, but still raise concerns about the current version that prevent it from being published in eLife. The reviewing editor has now carefully examined these remaining concerns including a review of the literature regarding in situ analysis of protein interactions using microcontact printing.

Essential revisions:

1) The current application of the method is thought to be of high value and provides significant biological insight. The method itself is too conceptually similar to earlier work from the Schütz lab (which you have cited correctly) to be referred to as a "novel" assay and thus it is requested that you remove the word "novel" from the title and anywhere else in the manuscript this is claimed, unless you specifically identify the novel aspect. The reviewers acknowledge that you have literally created a "two-hybrid" version based on using the epitope tagged bait (hybrid 1) and the fluorescent tagged prey (hybrid 2), which Schütz didn't fully realise, but they had already used the bait-prey terminology in their 2010 Jove paper, and conceptually it is too small an advance in terms of assay technology. They didn't need to make a first hybrid as they tested a transmembrane bait interacting with a cytoplasmic prey. But they have everything else you describe and more sophisticated analysis.

2) While the reviewers appreciate that you define cluster in the context of this study, "cluster" with respect to MHC class I antigens has been previously defined and this needs to be respected within papers in this field to avoid confusion. It is also felt that with the ability to detect only a 10% steady state association with the current method (based on use of 1.1 as a threshold) that you apparently can't detect the earlier defined MHC class I clusters. But you are not saying these don't exist. A compromise would be to refer to the phenomenon you are studying as "induced clusters" throughout to distinguish them from previously described spontaneous clusters of MHC class I trimers, that you neglect due to your detection threshold. It is essential to clarify this or use a term not including "cluster".

3) Proposed new title: "A two-hybrid antibody micropattern assay reveals conformation-specific cell surface induced clustering of MHC I proteins".

4) You describe the quantitative threshold of 1.1 in the response to reviewers, but we couldn't find this in the paper. Please include in the Materials and methods the 1.1 threshold for defining an induced cluster in the system. It is not clear how the gap between 1.1 and 1.3 is defined as a grey zone in the rebuttal, but you should also mention this if you feel it is important to actually set the threshold at 1.3, although this is very close to your experimental result.

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Thank you for choosing to send your work entitled "A novel two-hybrid antibody micropattern assay reveals cell surface clustering of MHC I heavy chains" for consideration at eLife. Your article has been reviewed by a Senior Editor and a Reviewing editor, and we are prepared to consider a revised submission with no guarantees of acceptance.

There is no editorial policy against use of "novel" in titles of eLife papers, but this has been infrequent – at around 0.2%. In the first decision letter the editors did override one reviewer in acknowledging that the approach was novel, but when the reviewer persisted in questioning this an editor went to the primary literature and looked at the precedents specifically to address this issue independently. As described in the second decision letter the opinion of the editors was then changed to agree with the reviewer that the core of the assay technology was well described earlier and that this assay is a novel application of an existing assay concept. Since this is too much information to convey in the title, the decision was to remove "novel" from the title and then explain the more nuanced novelty of this application in the text. The editors still ask that you remove novel from the title.

The second issue has to do with the definition of cluster and whether this is misleading in this context. The editors still ask that you change this terminology to avoid confusion and quantitative implications of the term. The interaction leading to a signal in this assay could be sub-stoichiometric, 1:1 or super-stiochiometric. The term cluster suggests groupings of greater than 1:1, because 1:1 would be called a dimer, and less than 1:1 would not be cluster. This method doesn't provide any information about stoichiometry, but just that it leads to co-localization of prey with the bait by fluorescence microscopy. The heavy chain interaction you are detecting may very well be a dimer, based on other literature, but this cannot be determined from this data. So describing the process as a cluster seems to be an over-interpretation. The bait and pray are simply co-localized at conventional visible light resolution. The authors should refer to co-localization, co-capture or enrichment to precisely describe what the assay detects as a ratio of on/off pattern as quantified in Figure 3?

The last issue was the baseline co-localization of prey with bait, which is described as generating a ratio of 1.1 of on/off pattern. This ratio was 1.3 for conditions leading to the empty MHC I heavy chains. Is 1.1 measured under conditions favoring the full heterotrimer of MHC I, peptide and β2m, significantly different than 1.0? The quantification was new data added by the authors to Figure 3. This is where 10% came from. The authors may not have data to determine if 1.1 is significantly different than 1.0, but as a discussion point, they might discuss that the previously describe interactions of MHC I proteins on cells may be captured in this number, which would then be a biological background for this assay method and thus will impact its sensitivity. This would also put the relative magnitude of other forms of MHC clustering and this phenomenon in context. More experiments would be needed to sort this out- probably use of different controls to find proteins that are totally unaffected by capture of some of the MHC I molecules to the patterned antibodies. The authors are urged to put these aspects of this assay and earlier investigations of MHC I distribution on the cell surface in context as best they can.

https://doi.org/10.7554/eLife.34150.013

Author response

Essential revisions:

1) Quantify imaging results and validate for this example. Given this signal to noise, how weak an association could be detected?

We thank the reviewer for this important comment. In our revised manuscript, we have now included the quantification of Kb-Kb cluster formation at 37 °C and the respective controls according to our observations in Figure 3B. We have thus added the quantification to Figure 3 (see new Figure 3C) and have extended the Results section (subsection “Antibody micropatterns reveal conformation-dependent in cis interactions of Kb free heavy chains”, fifth paragraph) as well as the Materials and methods section accordingly (subsection “Quantification of clustering”).

Briefly, for the quantification we used ImageJ to compare the mean fluorescence intensity of the entire cell with the mean fluorescence intensity of Kb-GFP in the areas of the pattern elements. A relocalization of Kb-GFP to the pattern elements leads to increased fluorescence intensity levels on the pattern elements and an increase of the ratio of the fluorescence intensity of the pattern elements over the fluorescence intensity of the entire cell. According to theoretical considerations and the observations in our system, we interpret a ratio of 1.3 as clustering and values below 1.1 as background signal (Figure 3C). Quantification is not trivial in our example since the Kb-transfected TAP-deficient STF1 cells have a generally strong ER background. This is because many Kb molecules that cannot be loaded with intracellular peptides are not transported to the plasma membrane. Despite these high background levels, we obtained significant differences between clustering cells and controls (see Figure 3C).

We think that the numerical fluorescence ratios might be different for other protein-protein interactions, depending on cell type, expression levels, background signal, and also the pattern element sizes. These factors will play important roles for the sensitivity of the assay but can be adjusted during optimization experiments.

Regarding the detection of weak protein-protein interactions asked by the reviewer, we think that our method can be even advantageous over conventional methods, since it allows for the accumulation of proteins to the pattern elements over time to increase the signal intensity. Also, one can increase the number of captured proteins by increasing the antibody concentration or achieve better contrast by varying pattern geometries and/or pattern element sizes. Alternatively, a readout by TIRF or super-resolution microscopy techniques might allow researchers to detect weak signals when only shorter incubation times on the antibody micropatterns can be realized, or when protein-protein interactions require temporal resolution.

2) We would strongly recommend not using the word cluster to describe the patterning of class I molecules with glass bound Abs. While this is clearly a useful technique, it is highly artificial, and for all we know, each "cluster" could be a single pair of co-localized Abs. Cluster has been used (appropriately) in the past to refer to spontaneously concentrated class I molecules, likely present in membrane domains that limit their ability to diffuse from each other. The use of cluster in the present study muddies the waters considerably, and will likely increase confusion, rather than the opposite.

We thank the reviewer for this comment, which has revealed to us an important potential misunderstanding. To understand our assay, our readers need to be able to discriminate between two events that we have termed captureand clustering. We still believe that this terminology is precise and useful according to the following definitions:

1) We use the term capture to describe the binding of cell surface membrane proteins to the antibody micropatterns. Capture thus describes the direct antibody-antigen interaction that prevents internalization of the bound proteins, i.e., the mechanical trapping of Kb proteins at the cell surface (see Figure 3).

2) In contrast, we use the term clustering to describe the binding of further protein molecules to the already captured proteins on the pattern elements. This is the event that our assay is set up to detect. This definition matches the statement of the reviewer: “Cluster has been used (appropriately) in the past to refer to spontaneously concentrated class I molecules, likely present in membrane domains that limit their ability to diffuse from each other.”

It is essential that the two terms are clearly introduced, differentiated, and understood by the readers. We very much agree that the term clustering might be misleading if not properly defined, but we are convinced that clustering is still the best term for our observation. We have therefore introduced a clearer definition of both terms in the manuscript (Introduction, second and third paragraphs). In addition, we have gone through our Introduction to make sure that the term capture is now stringently used.

With these adjustments, we now trust that our explanations are clear and straightforward, and that the readers will be able to distinguish between the two terms.

3) Is it possible to reverse the association of non-conformed Kb molecules by addition peptide and or β2m?

We thank the reviewer for pointing out this exciting possibility. Indeed, during our work, we have thought hard about the molecular structure of the clustered proteins, and especially if they would be able to bind peptide and/or β2m.

First, we tested peptide binding to the clustered Kb molecules in order to characterize the clustered Kb molecules in more detail (see Figure 3—figure supplement 2). In these experiments, we observed that peptide binding was impaired, suggesting that the clustered Kb molecules are no longer peptide-receptive. This is plausible since MHC class I free heavy chains do not usually bind peptide well.

Second, with respect to β2m binding, we have – in previous published work – performed β2m incubation experiments in the same experimental system of Kb in STF1 cells and found that exogenous β2m does not significantly increase the lifetime of surface class I, suggesting that it will not significantly re-bind to the Kb free heavy chain once the original β2m has dissociated (Montealegre et al., 2015).

In the work presented in this manuscript, we show that incubation with peptide prevents the formation of clusters (Figure 3B, column 4). We have not incubated the cells with peptides or with exogenous β2m after cluster formation in order to try to reverse the Kb-Kb interaction, as the reviewer suggested, but from the above observations of impaired peptide and β2m binding, we assume that Kb free heavy chain clusters will not be dispersed by the addition of β2m or peptide.

We believe that the detailed molecular mechanism of cluster formation, its potential dynamic nature, and the subsequent fate of the class I free heavy chain clusters are very interesting questions to follow up on, but we would like to leave these investigations to a future manuscript.

4) Subsection “Stabilizing effect of conformation-specific antibodies allows for differential patterning of Kb dimers and free heavy chains”, first paragraph: regarding the use of TAP negative cells (whose original publication should be cited), Day et al., 1994, reported that TAP positive cells actually express more peptide receptive cell surface molecules, undermining the reasoning of these statements.

We thank the reviewer for carefully reading our manuscript. In our revised manuscript we have now added the original publication of STF1 cells (de la Salle et al., 1999) specifically to the mentioned sections as requested. We have also cited the Day et al., 1995 paper.

As the reviewer commented correctly, Day and collaborators investigated the role of TAP for the generation of peptide-receptive molecules at the cell surface. In their work, they found that the murine TAP-positive RMA cells have higher surface levels of peptide-receptive Kb molecules than TAP-deficient RMA-S cells (Day et al., 1995).

For our studies, we opted for a human TAP2-deficient cell line and transduced it with the Kb constructs described in the manuscript in order to bypass any murine MHC class I background issues in our system. Prompted by the question of the reviewer, we became curious as to the levels of peptide-receptive Kb molecules on the surface of the STF1 cells, and similar to the work of Day et al., we have now compared STF1 cells with STF1+TAP2 cells (STF1 cells transduced with TAP2, which have wild type TAP function). We transduced both cell lines with HA-Kb, thus generating the two stable cell lines (STF1/HA-Kb and STF1+TAP2/HA-Kb). To quantify peptide-receptive molecules, we added the Kb-specific high-affinity peptide SIINFEKL (SL8; 10 µM) to the cell culture medium and incubated the cells for 20 min at 37 °C. The cells were then washed, trypsinized, and stained with antibodies according to the standard protocol for flow cytometry.

Author response image 1
Comparison of cell surface expression of HA-Kb in STF1 and STF1+TAP2 cells by flow cytometry.

STF1 cells were transduced with HA-Kb and stained with anti-HA or 25-D1.16 and anti-mouse IgG conjugated to Alexa Fluor 488 and subjected to flow cytometry. (A) Surface intensities of HA-Kb of both cell lines are represented as bar charts. TAP-deficient STF1 cells are represented in black and TAP2-proficient cells (STF1+TAP2) are represented in grey. Cells were incubated with (+) and without (-) 10 µM of the high affinity peptide SL8 and stained with the indicated antibodies. (B) The increase in surface signal after peptide addition in (A) for both antibodies is displayed as ratio. (n= 3; standard deviations as indicated).

https://doi.org/10.7554/eLife.34150.012

The total Kb surface levels were determined by staining with the anti-HA antibody (Author response image 1A). In STF1 cells, Kb surface levels only showed a small and barely significant increase upon addition of SL8, probably because the peptide prevented endocytosis of some empty dimers (Montealegre et al., 2015). The Kb surface levels in STF1+TAP2 cells remained the same upon peptide addition, although they have generally higher Kb surface levels (Author response image 1A).

In order to determine the peptide-receptive population of the detected Kb molecules, the same samples were stained with the 25-D1.16 antibody, specifically binding to Kb molecules that are loaded with the SL8 peptide (Porgador et al., 1997). As expected, the 25-D1.16 signal increases clearly upon addition of the SL8 peptide (see Author response image 1B).

Remarkably, the almost identical increase of the 25-D1.16 signal after peptide addition shows that the levels of peptide-receptive molecules are similar in both cell lines. This suggests that in both cell lines, despite their very different Kb surface levels (Author response image 1A), similar mechanisms are at work to limit the amount of peptide-receptive ('empty') class I at the cell surface.

These findings are in contrast to the findings of Day and collaborators on RMA-S cells that the reviewer mentions. We are not surprised, though, since class I quality control and trafficking parameters vary between different cells; one important difference is that RMA-S cells are lymphocytes, in which the antigen processing and presentation genes in the MHC are fully induced (whereas STF1 cells are fibroblasts).

Our original reason for using TAP-deficient STF1 cells have to do with the homogeneity of the class I population. The experiments in our manuscript rely on our ability to generate free heavy chains of H-2Kb on the cell surface, and to trap them there. This is achieved by capturing the Kb heavy chain/β2m dimers and incubating the cells at 37 °C to allow these dimers to dissociate, leaving captured free heavy chain that is unable to bind peptide (see answer to comment 3 above, and Figure 2B, column 3, of the manuscript). These free heavy chains then cluster with other free heavy chains (Figure 3B, column 3). Since we use TAP-deficient STF1 cells, the majority of the Kb molecules that reach the surface are probably heavy chain/β2m dimers, which provides us with a relatively homogeneous population. If, according to the question of the reviewer, we were to use STF1+TAP2 cells, then we would additionally have a large population of Kb heavy chain/β2m/peptide trimers (compare the STF1 and the STF1+TAP2 columns in Author response image 1A), which do not turn into free heavy chains at 37 °C and thus provide a second species of Kb at the cell surface that does not undergo clustering and possibly dilutes our readout signal. This, for us, was the main reason for using STF1 cells for the experiments.

We have not made any changes to the manuscript as a result of the work described in this answer.

5) "So their endogenous class I molecules would not interfere with the Ab micropatterns". This should be rephrased as it is not clear.

We thank the reviewer for carefully reading our manuscript and for the suggested corrections. We have changed the indicated sentences in the revised manuscript accordingly (see subsection “Cell lines and gene expression”).

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Essential revisions:

1) The current application of the method is thought to be of high value and provides significant biological insight. The method itself is too conceptually similar to earlier work from the Schütz lab (which you have cited correctly) to be referred to as a "novel" assay and thus it is requested that you remove the word "novel" from the title and anywhere else in the manuscript this is claimed, unless you specifically identify the novel aspect. The reviewers acknowledge that you have literally created a "two-hybrid" version based on using the epitope tagged bait (hybrid 1) and the fluorescent tagged prey (hybrid 2), which Schütz didn't fully realise, but they had already used the bait-prey terminology in their 2010 Jove paper, and conceptually it is too small an advance in terms of assay technology. They didn't need to make a first hybrid as they tested a transmembrane bait interacting with a cytoplasmic prey. But they have everything else you describe and more sophisticated analysis.

This is for us the first point of confusion. Perhaps the Editor would like to explain. In the comments on our original submission, the Editors wrote:

"The reviewers agreed that the use of micro contact printing to evaluate lateral interactions MHC class I proteins in situ was novel and presented a useful new approach to such problems." Based on this statement, we assumed that the use of the word 'novel' was supported by reviewers and Editors. Does the more recent comment now mirror the opinion of additional reviewers? Or did the original reviewers change their minds?

In our own opinion, the novelty of our approach is threefold: first, the simple technique of the antibody micropatterns, which in principle allows the method to be adapted by any molecular immunology or cell biology laboratory; second, the ability to investigate, by use of the anti-tag antibody, a defined conformation of MHC class I (and not just any and all MHC class I protein; we are not aware of any other technique that can differentiate different protein conformations in such a clear fashion); and third, the use of the technology to identify and characterize a previously uncharacterized (of course, class I homotypic interactions were previously found in the Zuñiga paper, but it was not possible to her and her coworkers to determine that only the heavy chains were interacting; see the Discussion) interaction (to our knowledge, all previous applications of micropatterns were just proof-of-principle applications, common in biophysical publications, of interactions that were already well-characterized by cell biologists).

We have left the word 'novel' standing in the manuscript for now, but we will not insist on it if the Editor believes that novelty does not apply. We think that upon publication, the quality and novelty of the work will speak for itself; of course, we think that an article will attract more readers and ultimately citations if the word 'novel' is used wherever justified.

2) While the reviewers appreciate that you define cluster in the context of this study, "cluster" with respect to MHC class I antigens has been previously defined and this needs to be respected within papers in this field to avoid confusion. It is also felt that with the ability to detect only a 10% steady state association with the current method (based on use of 1.1 as a threshold) that you apparently can't detect the earlier defined MHC class I clusters. But you are not saying these don't exist. A compromise would be to refer to the phenomenon you are studying as "induced clusters" throughout to distinguish them from previously described spontaneous clusters of MHC class I trimers, that you neglect due to your detection threshold. It is essential to clarify this or use a term not including "cluster".

2.1) About the use of the word 'clusters'. While we 100% appreciate that confusion must be avoided, we think that the word 'cluster' has no clear definition in the field, but has actually been used in three completely different contexts in the literature. We have seen:

- Lu et al., 2012, who define clusters as protein islands on the cell surface. In their paper, they found that certain pMHC (Kb+SIINFEKL or Kb+ SIYRYYGL) are found in distinct areas or “clusters” (200-900 nm in diameter). They conclude that “Our most important finding is that endogenous antigen processing generates intracellular clusters of class I molecules segregated on the basis of their peptide cargo that are maintained for hours after their delivery to the cell surface.”

- Pentcheva and Edidin, 2001 (Pentcheva, T., and M. Edidin. 2001. Clustering of peptide-loaded MHC class I molecules for endoplasmic reticulum export imaged by fluorescence resonance energy transfer. J. Immunol. 166:6625–32. doi:10.4049/JIMMUNOL.166.11.6625), who see associations ('clusters') of A2 in the ER;

- Mocsár et al., 2016, who have seen MHC I-Interleukin Clusters (including all MHC I isotypes: A, B, C).

In the review of our first submission, the following sentence is found:

"Cluster has been used (appropriately) in the past to refer to spontaneously concentrated class I molecules, likely present in membrane domains that limit their ability to diffuse from each other. The use of cluster in the present study muddies the waters considerably, and will likely increase confusion, rather than the opposite." We understand from this that the reviewer(s) refer(s) to the Edidin clusters, which are speculated to have something to do with lipid environments. If that is not the case, we would ask the Editor to point out to us the definition of 'clusters' in the literature that they find primary and conclusive. Perhaps, for clarification, it would have been useful if we had been able to see the original remarks of the reviewer.

In addition to the three different definitions of MHC clusters in the literature, the word 'cluster' is also a commonsense word in the English language, and this is the sense that we would have liked to use. It is important to use a commonsense word since people use search engines such as PubMed to find scientific work, and the words that are used to describe phenomena must be commonsense, so that people will be able to identify our paper in eLife. This speaks strongly against the use of unusual or invented terms.

Finally the Editor suggests the term 'induced clusters', which we find not fitting, since there is nothing induced or inducible in our associations of class I molecules. One possibility, from our point of view, is 'free heavy chain clusters', which also clearly differentiates from the Edidin work.

The Editor is, and the reviewers are, invited to suggest an alternative wording that is commonsense, correct, and findable. For now, we have left the word 'cluster' standing in the manuscript.

It is also felt that with the ability to detect only a 10% steady state association with the current method (based on use of 1.1 as a threshold) that you apparently can't detect the earlier defined MHC class I clusters.

2.2) It is unclear to us how the reviewer arrived at these two statements, namely that s/he believes that the steady-state association that we show is only 10% (of what?) and that s/he believes that we are somehow not detecting other associations that are described in the literature. We are puzzled as to how a 1.1 threshold would correlate with a '10% steady-state association'. Again, perhaps it would have been helpful if we had been able to read the reviewer's original comment, since there was clearly a misunderstanding which must now be addressed.

In the following, we describe again our method for quantifying the conglomerates as already detailed in our Materials and methods section and in the earlier Smallpaper. To compare the spatial distribution of Kb-GFP between the pattern elements and the pattern element interspaces, we proceeded as follows:

1) We determined, for each image, a) the mean fluorescence intensity of the entire cell area, and b) the mean fluorescence intensity of the areas on the pattern elements. Quantification was done with ImageJ (National Institutes of Health, Bethesda, USA).

2) We calculated, for each image, the ratio of these numbers of the mean fluorescence intensity on pattern elements over total fluorescence intensity. This means: a theoretical ratio of 1.0 describes a homogeneous distribution of proteins where the pattern elements and the interspaces have the same mean fluorescence intensity.

3) We then checked this ratio in our control experiments, where no clustering was expected, and it was 1.1. This value was thus defined as the threshold.

4) In the experiments, a clear redistribution of proteins onto the pattern elements corresponds to a ratio of 1.3, with the appropriate significance.

With this type of evaluation, we were not at all aiming to determine the percentage of interacting proteins, nor does this analysis allow us to derive them.

It is also felt that with the ability to detect only a 10% steady state association with the current method (based on use of 1.1 as a threshold) that you apparently can't detect the earlier defined MHC class I clusters. But you are not saying these don't exist.

2.3) We think that the reviewer is referring to the MHC class I clusters defined by Yewdell or Edidin (see 2.1.). The fundamental difference between our clusters and these is the following: our clusters are clearly made up of free heavy chains, whereas theirs are associations of trimers (trimer: complex of heavy chain, light chain, and peptide) or mixed associations of trimers and heavy chain/β2m dimers.

We do not say that the Yewdell or Edidin clusters do not exist, since we are not in a position to detect them. We are using TAP-deficient cells, in which very few peptides are available to the class I molecules, and so clusters that consist of, or that contain, trimers are not visible in our experimental system. It is important for the reviewer to appreciate this difference. We have now added a sentence in the Discussion that clearly differentiates our clusters from the Yewdell and Edidin clusters.

A compromise would be to refer to the phenomenon you are studying as "induced clusters" throughout to distinguish them from previously described spontaneous clusters of MHC class I trimers, that you neglect due to your detection threshold.

2.4.With respect to the terms 'cluster' and 'induced', please see our statements above in 2.1.

With respect to the 'detection threshold', please see our statements above in 2.2.

3) Proposed new title: "A two-hybrid antibody micropattern assay reveals conformation-specific cell surface induced clustering of MHC I proteins".

We think that it is not prudent to use the word 'induced' (see above 2.1.). One possible similar title would be "A novel two-hybrid antibody micropattern assay reveals cell surface clustering of MHC I heavy chains". We have changed the title of the manuscript accordingly.

4) You describe the quantitative threshold of 1.1 in the response to reviewers, but we couldn't find this in the paper. Please include in the Materials and methods the 1.1 threshold for defining an induced cluster in the system.

The method was referred to from the Small paper (Dirscherl et al., 2017). We have now added the definition and explanation to the Materials and methods in this manuscript as requested.

It is not clear how the gap between 1.1 and 1.3 is defined as a grey zone in the rebuttal, but you should also mention this if you feel it is important to actually set the threshold at 1.3, although this is very close to your experimental result.

We have now added the explanation to the Materials and methods.

[Editors' note: further revisions were requested prior to acceptance, as described below.]

There is no editorial policy against use of "novel" in titles of eLife papers, but this has been infrequent – at around 0.2%. In the first decision letter the editors did override one reviewer in acknowledging that the approach was novel, but when the reviewer persisted in questioning this an editor went to the primary literature and looked at the precedents specifically to address this issue independently. As described in the second decision letter the opinion of the editors was then changed to agree with the reviewer that the core of the assay technology was well described earlier and that this assay is a novel application of an existing assay concept. Since this is too much information to convey in the title, the decision was to remove "novel" from the title and then explain the more nuanced novelty of this application in the text. The editors still ask that you remove novel from the title.

We have now removed the term “novel” from our title.

The second issue has to do with the definition of cluster and whether this is misleading in this context. The editors still ask that you change this terminology to avoid confusion and quantitative implications of the term. The interaction leading to a signal in this assay could be sub-stoichiometric, 1:1 or super-stiochiometric. The term cluster suggests groupings of greater than 1:1, because 1:1 would be called a dimer, and less than 1:1 would not be cluster. This method doesn't provide any information about stoichiometry, but just that it leads to co-localization of prey with the bait by fluorescence microscopy. The heavy chain interaction you are detecting may very well be a dimer, based on other literature, but this cannot be determined from this data. So describing the process as a cluster seems to be an over-interpretation. The bait and pray are simply co-localized at conventional visible light resolution. The authors should refer to co-localization, co-capture or enrichment to precisely describe what the assay detects as a ratio of on/off pattern as quantified in Figure 3?

We have now changed our manuscript and avoid the term 'cluster' as requested. Instead, we describe our findings with the terms “in cis (protein-protein) interactions”, “associated” or “co-captured” MHC I proteins. When referring to the findings of other groups, we use the terms that they have used to describe their observations, including the term 'clusters'.

The last issue was the baseline co-localization of prey with bait, which is described as generating a ratio of 1.1 of on/off pattern. This ratio was 1.3 for conditions leading to the empty MHC I heavy chains. Is 1.1 measured under conditions favoring the full heterotrimer of MHC I, peptide and β2m, significantly different than 1.0? The quantification was new data added by the authors to Figure 3. This is where 10% came from. The authors may not have data to determine if 1.1 is significantly different than 1.0, but as a discussion point, they might discuss that the previously describe interactions of MHC I proteins on cells may be captured in this number, which would then be a biological background for this assay method and thus will impact its sensitivity. This would also put the relative magnitude of other forms of MHC clustering and this phenomenon in context. More experiments would be needed to sort this out- probably use of different controls to find proteins that are totally unaffected by capture of some of the MHC I molecules to the patterned antibodies. The authors are urged to put these aspects of this assay and earlier investigations of MHC I distribution on the cell surface in context as best they can.

We have now extended our discussions and included the points raised by the Editors.

https://doi.org/10.7554/eLife.34150.014

Article and author information

Author details

  1. Cindy Dirscherl

    Department of Life Sciences and Chemistry, Jacobs University, Bremen, Germany
    Contribution
    Conceptualization, Supervision, Validation, Investigation, Visualization, Methodology, Writing—original draft
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-8973-2835
  2. Zeynep Hein

    Department of Life Sciences and Chemistry, Jacobs University, Bremen, Germany
    Contribution
    Conceptualization, Investigation, Writing—review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-6335-8961
  3. Venkat Raman Ramnarayan

    Department of Life Sciences and Chemistry, Jacobs University, Bremen, Germany
    Contribution
    Investigation, Acquisition of data, Analysis, Interpretation of data
    Competing interests
    No competing interests declared
  4. Catherine Jacob-Dolan

    Department of Life Sciences and Chemistry, Jacobs University, Bremen, Germany
    Contribution
    Investigation, Acquisition of data, Analysis, Interpretation of data
    Competing interests
    No competing interests declared
  5. Sebastian Springer

    Department of Life Sciences and Chemistry, Jacobs University, Bremen, Germany
    Contribution
    Conceptualization, Supervision, Funding acquisition, Project administration, Writing—review and editing
    For correspondence
    s.springer@jacobs-university.de
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5527-6149

Funding

Deutsche Forschungsgemeinschaft (SP583/7-2)

  • Sebastian Springer

Tönjes Vagt Foundation (XXXII)

  • Sebastian Springer

Bundesministerium für Bildung und Forschung (031A153A)

  • Sebastian Springer

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

The authors thank Susanne Illenberger und Susanne Fritzsche for suggestions on the manuscript; and Ursula Wellbrock for the cultivation of hybridoma cell lines and excellent technical assistance.

Senior Editor

  1. Arup K Chakraborty, Massachusetts Institute of Technology, United States

Reviewing Editor

  1. Michael L Dustin, University of Oxford, United Kingdom

Publication history

  1. Received: December 7, 2017
  2. Accepted: July 24, 2018
  3. Version of Record published: September 5, 2018 (version 1)

Copyright

© 2018, Dirscherl et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 496
    Page views
  • 56
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)