Secreted and Transmembrane 1a Is a Novel Co- Stimulatory Ligand

Most T cell responses to pathogens or self antigens are modulated through the action of regulatory T cells and tissue-specific inhibitory mechanisms. To this end, several receptor-ligand pairs have evolved which either augment or diminish T cell function. Here we describe the tissue ligand SECTM1A (Secreted and transmembrane1A) as an alternative murine CD7 ligand. We show that SECTM1A, like SECTM1B, binds strongly to CD7, and that SECTM1B was able to compete with SECTM1A for CD7 binding. SECTM1A is ubiquitously expressed and has two major alternative transcripts which differ in expression between tissues. Both immobilised soluble forms of SECTM1A and SECTM1B and cell surface anchored forms demonstrated opposing effects on CD4+ T cell activation. Whereas SECTM1A acted as a co-stimulator of T cells, enhancing IL-2 production and proliferation, SECTM1B proved inhibitory to TCR mediated T cell activation. Surprisingly, both functional outcomes proved to be CD7-independent, indicating the existence of alternative receptors for both ligands. We used a SECTM1A-Fc fusion protein to immunoprecipitate potential alternative ligands from detergent lysates of CD7 2/2 T cells and, using mass spectrometry, identified GITR as a SECTM1A binder. SECTM1A was found to bind to activated CD4+ and CD8+ T cells as well as to CHO cells expressing cell surface GITR. Binding of SECTM1A to activated primary T cells was inhibited by either GITRL-Fc or anti GITR antibodies. Thus SECTM1A and SECTM1B represent novel reciprocal alternative ligands which may function to modulate the activation of effector and regulatory T cells. The ability of SECTM1A to activate T cells may be explained by its ability to bind to GITR. Copyright: ß 2013 Howie et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Funding: This work was funded by two programme grants from the Medical Research Council www.mrc.ac.uk G0400808 (MHB) and G7904009 (HW). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.


Introduction
T cell responses to antigen are controlled at multiple levels. These include control of the cells' anatomical location, local MHC-antigen concentration on APCs, intracellular signalling changes, inhibition by regulatory T cells and soluble and cell surface regulatory ligand molecules expressed by antigen presenting cells. Examples of the latter group include receptor/ligand pairs such as GITR and GITR ligand, previously described by our laboratory to control activation of effector and regulatory T cells [1] PD1/PDL, CTLA4/B71&2, 41BB/41BBL, ICOS/ICOS ligand and CD7/SECTM1B (Secreted and transmembrane1B). CD7 is a T cell and NK cell expressed 40 kD transmembrane protein of the immunoglobulin superfamily which can induce lipid and tyrosine kinase activity and is thought to be important for T/ NK cell activation [2].
A report of aberrant Treg development and function in CD7/ CD28 double knockout mice [3], and our observation of elevated CD7 expression by TGFb-induced regulatory T cells (unpublished observations) prompted us to investigate the effects of known CD7 ligands on T cell responses. Human and murine CD7 have previously been shown to interact with cell surface and secreted members of the immunoglobulin superfamily named SECTM1 (K12) and SECTM1B respectively [4].
Murine SECTM1B identified by Lyman et al [4] had been shown to inhibit proliferation of ConA-stimulated T lymphocytes whilst up regulating activation markers on activated NK cells. The human genome contains only one SECTM gene, SECTM1 which is situated in close proximity to CD7 on chromosome 17q25. As described here the mouse genome contains two SECTM genes in close proximity to the CD7 gene. Mouse SECTM1A shares the greatest homology to human SECTM1. SECTM1-Fc fusion proteins were shown by Lyman and colleagues to increase surface activation markers on human NK cells [4]. Wang et al reported SECTM1-Fc fusion proteins co-stimulated IL-2 and interferon gamma production by human CD4+ and CD8+ cells [5]. SECTM1 is predicted to be a type 1A transmembrane protein with extracellular N terminus and cleaved signal peptide. Depending on the experimental system used, SECTM1 appears either as a Golgi-associated protein which is secreted as a 20 kDa form, in breast carcinoma cells [6] or as a plasma membrane expressed surface protein in COS-1 cells transfected with expression plasmids [4]. SECTM1 is predominantly expressed by cells of the myeloid lineage and epithelia where its expression is enhanced by exposure to interferon gamma in a Stat-1 dependent manner [7]. It is possible that its ultimate location, secreted or on the plasma membrane, is dependent on either the cell on which it is expressed, the tissue location or micro environmental factors such as inflammation.
In this study we report the function of the novel murine CD7 ligand SECTM1A whose gene is located in the CD7/SECTM1B locus on mouse chromosome 11 with a tissue distribution differing from the known CD7 ligand SECTM1B. We constructed fusion proteins between the extracellular domains of SECTM1A and SECTM1B with an aglycosyl mutant Fc region of human IgG1, or anchored forms expressed on the cell surface of CHO cells. These were used to probe the function of these ligands in the experiments described here. Exposure of T cells to SECTM1A or SECTM1B resulted in reciprocal modulation of T cell proliferation. Surprisingly, despite SECTM1A being a ligand for CD7, all of these signalling events turned out to be CD7 independent. The costimulatory functions of SECTM1A may be explained in part by our observation that it is able to interact, albeit weakly with GITR as an alternative receptor.

Results
The identification of mSECTM1A as a novel mCD7 ligand with homology to hSECTM1 and mSECTM1B The gene previously described as the murine orthologue of human SECTM1, murine Sectm1b [4], and Sectm1a, a newly annotated gene encoding an uncharacterised protein, lie in a 44kbp region neighbouring CD7 ( Figure 1A). Alignment of the amino acid sequences of SECTM1A, SECTM1B and human SECTM1 reveals moderate homology between the three proteins in their extracellular regions, the greatest homology being between mouse SECTM1A and human SECTM1 which share 43% amino acid identity ( Figure 1B). Cloning of Sectm1a from splenic cDNA revealed a major splice variant lacking exon IV (data not shown). Exon IV encodes a stalk region joining the immunoglobulin domain to the transmembrane domain ( Figure 1C). Tissue distribution of Sectm1a, 1a-ExIV and Sectm1b transcripts was assessed by real-time RT-PCR ( Figure 1D-F). Primers and probes used for this analysis are shown diagrammatically in figure 1C. Sectm1b had a variable distribution with the highest expression detectable in large and small intestine and being undetectable in bone marrow and lymph nodes ( Figure 1D). Primers spanning exons 3 and 4 were used to measure full length Sectm1a only ( Figure 1E) whereas those spanning exons 3-5 detect 1a-ExIV ( Figure 1F). Sectm1a and 1a-Exon IV were more widely expressed than sectm1b.
To probe the functions of murine SECTM1A/B we constructed fusion proteins between the extracellular domains of SECTM1A, SECTM1A-ExIV, SECTM1B and CD7 with the aglycosyl mutated Fc portion of human IgG1. Full length SECTM1A-Fc fusion proteins were relatively labile therefore SECTM1A-ExIV was used in most subsequent investigations (termed ''SECTM1A'' from now on). SECTM1A, SECTM1B and CD7, are predicted to have 4, 3 and 3 N-linked glycosylation sites respectively. Digestion with peptide N-glycosidase F revealed a shift in migration by SDS PAGE confirming that the Fc-fusion proteins were indeed glycosylated ( Figure 2A).
We used surface plasmon resonance to test the functionality of the SECTM fusion proteins and to test the hypothesis that SECTM1A might be an alternative ligand for CD7. CD7, SECTM1A and SECTM1B-Fc-fusion proteins were immobilised onto BIAcoreTM CM5 chips and the same proteins were injected over the flow cells in solution to test binding. CD7-Fc in solution caused a rapid increase in response units indicating fast binding to both SECTM1A and SECTM1B in a reproducible manner (bold solid and dotted lines, Figure 2B &E). Thus SECTM1A represents a novel alternative ligand for mouse CD7.
CD7-Fc dissociated from SECTM1A more rapidly than SECTM1B suggesting that CD7 binds with higher affinity to SECTM1B than SECTM1A ( Figure 2B and 2E, dissociation at 8-9 minutes bold solid and dotted lines). We then tested whether SECTM1B could block binding of SECTM1A to CD7. SECTM1B-Fc was passed over solid phase CD7 and bound rapidly ( Figure 2C, solid line). When SECTM1A was passed over the same flow cell directly after SECTM1B no binding above background was observed ( Figure 2D, solid line). Thus SECTM1B competes for CD7 binding with SECTM1A. In a reciprocal experiment, SECTM1A was passed in solution over solid phase CD7 ( Figure 2F, solid line) and binding was observed albeit with less increase in response units seen for SECTM1B. Directly following SECTM1A, SECTM1B was passed over the same flow cell. This time binding to CD7-Fc was observed with almost exactly the same stoichiometry as when SECTM1B bound CD7-Fc alone ( Figure 2G, solid line) which is consistent with a weaker affinity of SECTM1A for CD7. SECTM1A and SECTM1B did not exhibit homophilic binding nor did they bind to each other under these conditions.

SECTM1B inhibits activation and IL-2 production of T cells
It is unclear what functions SECTM1A and SECTM1B serve. Our preliminary experiments with SECTM1B-Fc showed no evidence for modulation of T cell responses when the fusion protein was used in solution (data not shown). However solid phase SECTM1B-Fc but not CD7-Fc inhibited both IL-2 production and proliferation of T cells activated through anti-CD3 stimulation ( Figure 3 A & B). The inhibitory effect was not due to steric hindrance of the anti-CD3 antibodies as equal protein was loaded onto each culture well using Campath-1H (anti human CD52 with the same human IgG1, aglycosyl Fc fragment as the fusion proteins) as an irrelevant aglycosyl human IgG1 antibody 'filler'. The effects of SECTM1B were not due to induction of apoptosis as shown by trypan blue exclusion and annexin V staining (data not shown). Inhibition of T cell activation by SECTM1B occurred in the presence of co-stimulation and exogenous IL-2 ( Figure 3C) and was independent of TGFb as addition of neutralising anti-TGFb (1D11, [8]) did not reverse this effect. We then tested whether the mechanism of SECTM1B inhibition relied on soluble inhibitory factors being released into the cell culture supernatant. Culture supernatants from the experiment shown in Figure 3C were transferred to fresh cultures of anti-CD3 stimulated CD4+ T cells ( Figure 3D) showing equivalent proliferation of T cells grown for 48 hours in the presence of each supernatant. Thus this does not appear to be a major mechanism, however we cannot exclude the possibility that soluble inhibitory factors may be produced which immediately bind to the T cells, effectively sequestering them out of the culture supernatants.

SECTM1A is co-stimulatory for T cell activation
We compared the effects of SECTM1A and SECTM1B on T cell activation. First we compared the level of the activation marker CD69 expressed on the surface of CD4+ T cells following 48 hours of stimulation with immobilised anti-CD3 in the presence of immobilised SECTM1A or SECTM1B. Following activation in this manner SECTM1B reduced the percentage of cells expressing cell-surface CD69 ( Figure 3E). Surprisingly CD4+ T cells activated in the presence of SECTM1A showed an increase of CD69 over those cultured in the presence of the control protein.
Thus the cell surface phenotype of cells stimulated in the presence of SECTM1B or SECTM1A was consistent with inhibitory or costimulatory signalling respectively.
Both splice variants of SECTM1A, full length and lacking exon IV, enhanced T cell activation. CD4+ T cells activated in the presence of the full length ligand or the truncated isoform lacking exon IV also showed enhanced proliferation with increasing concentrations of plate-bound SECTM1A ( Figure 3F).
We then investigated whether SECTM1A and B are costimulatory or co-inhibitory respectively when situated on the plasma membrane of an antigen presenting cell. In order to reduce the contribution from other co-stimulation molecules we used CHO cells which were stably transfected with MHC class II (IE k ). We transfected these cells with membrane targeted constructs in the vector pDISPLAY ( Figure 4A) which targets Myc and HA tagged fusion proteins, the extracellular domains of SECTM1A, B or PDL-1 in this case, to the plasma membrane via the transmembrane domain of the platelet derived growth factor receptor (PDGFR, Figure 4B). CD4+ T cells from female A1.RAG 2/2 TCR transgenic mice, whose T cells recognise the male antigen, were cultured in the presence of CHO.IE k cells expressing the SECTM1 constructs pulsed with male antigen. Following 48 hours of culture A1.RAG 2/2 T cells stimulated in the presence of CHO cells expressing SECTM1A secreted significantly more IL-2 and those cultured in the presence of CHO cells expressing SECTM1B significantly less IL-2 ( Figure 4C). Thus SECTM1A and B represent co-stimulatory and inhibitory receptors in the context of MHCII/peptide mediated T cell activation.
One explanation for opposing functional outcomes from SECTM1A and B is that although both bind CD7, as detected by BIAcore, they may have additional alternative receptors on T cells. To test this hypothesis we used CD4+ T cells from C57BL/6 CD7 2/2 mice and found ( Figure 5) that remarkably, in in vitro stimulation assays, both SECTM1A-mediated co-stimulation and SECTM1B mediated inhibition of T cell proliferation occur in the absence of CD7. This indicates that there must indeed be additional receptors, other than CD7, for both these ligands on T cells.
To identify potential alternative SECTM1 ligands we first determined the cell types expressing the ligands. To this end we used allophycocyanin-labelled SECTM1A and SECTM1B Fc fusion proteins in FACS to stain CD7 2/2 resting and conA activated splenocytes. Despite its potent inhibitory effects on T cell activation in the solid phase, SECTM1B-Fc did not bind detectably to wt or CD7 2/2 resting or activated splenocytes in this FACS assay (data not shown). However SECTM1A-Fc was shown to bind to activated CD4+, CD8+ and to a lesser extent NK cells and granulocytes from CD7 2/2 mice ( Figure 6A).
As we were able to demonstrate binding of SECTM1A to activated CD7 2/2 CD4+ T cells by FACS, we used lauryl maltoside lysates of concanavalin A-activated CD7 2/2 splenocytes to immunoprecipitate the potential ligand(s) as described in methods. Eluates from SECTM1A-Fc/protein G agarose columns contained multiple proteins of different molecular weights (figure 6B). Coomassie stained gel slices containing the eluted proteins were tryptically digested and subjected to MS/MS mass spectroscopy for identification. Peptides with a MASCOT score greater than 90 are shown in figure 6C. The only protein with a plasma membrane transmembrane segment other than SECTM1A-Fc itself which was derived from the immunoprecipitation procedure was TNFRSF18 (GITR).
To confirm the SECTM1A/GITR interaction we tested whether binding of SECTM1A to activated CD7 2/2 CD4+ T cells could be inhibited by pre-incubation of the cells with either anti-GITR antibodies or GITRL. As shown in figure 6D and 6E both anti-GITR and soluble GITRL led to a reduced intensity staining of cells by SECTM1A-Fc. The specificity of the interaction was confirmed by utilising CHO cells expressing SECTM1A, SECTM1B, PDL-1 or GITR as targets for FACS staining with SECTM1A-Fc ( Figure 6F). Only CHO cells Alignment of Sectm1a, Sectm1b and human SECTM1 was performed in the Jalview program using the MUSCLE algorithm [31]. Shading indicates the degree of amino acid identity between the human SECTM1 and mouse SECTM1A and SECTM1B. Dark grey indicates a residue conserved in all three sequences. Light grey indicates a residue shared by only two sequences. No shading indicates areas of divergent sequence. The boxed region within the extracellular region of SECTM1A indicates the peptide encoded by exon IV C. Positions of the taqman primers and probes used to quantify expression of the different splice variants of SECTM1A. D. Tissue distribution of Sectm1b, E. Sectm1a (full length) F. Sectm1a -ExIV transcripts. mRNA was quantified using the taqman primers and probes illustrated in Figure 1A. Samples were normalised to the level of Sectm1b in brain which was given an arbitrary value of 1. doi:10.1371/journal.pone.0073610.g001

Discussion
In this study we have demonstrated a novel mode of immune regulation via the opposing actions of two murine immunoglobulin superfamily proteins SECTM1A and SECTM1B. These widely expressed plasma membrane proteins modulate activation of CD4+ T cells. We have demonstrated binding of both proteins to CD7 with surface plasmon resonance; however, they can exhibit in vitro functions independently of CD7. In the case of SECTM1A this may be explained by its ability to bind to GITR in the absence of CD7.
We observed significant differences in transcript expression between tissues. SECTM1B had a more restricted distribution than SECTM1A. SECTM1B transcripts were undetectable in bone marrow, heart, lymph nodes, brain, muscle, pancreas and skin whereas SECTM1A transcripts were detectable in these tissues although there were differences between SECTM1A and the spliced form lacking exon IV. It may be that that the presence of only SECTM1A in lymph nodes might help to promote T cell priming whereas the balance of SECTM1A/B in peripheral tissues could promote or inhibit T cell responses. Analysis of potential peptidase sites in the exon IV region of the SECTM1A protein reveals four potential cleavage consensus sites for glutamyl and aspartic endopeptidases which may generate soluble forms of the protein from the full length precursor (data not shown). It remains to be shown whether the spliced transcript of SECTM1A lacking exon IV could serve to provide a stable membrane-anchored form of the protein whilst the full length form could act as a precursor of a soluble form of the protein.
Our Biacore data clearly show that SECTM1B has a higher affinity for CD7 than SECTM1A and this hierarchy of binding may also extend to the alternative SECTM receptors on T cells. Differences in functional effects suggest that the differences in affinity for CD7 may apply to binding to these alternative receptors. Although we were unable to identify the alternative SECTM1B receptor in this study it is possible that both SECTM1A and B share other unidentified alternative receptors. An alternative explanation for the reciprocal effects of SECTM1A and B may be due to differences in efficacy of blocking rather than crosslinking of the same receptor. Monovalent binding by the higher affinity reagent SECTM1B-Fc would result in blocking whereas the lower affinity SECTM1A-Fc would be an effective crosslinking reagent resulting in activation. Similar effects of affinity and valency have been noted before and quantitatively analyzed [9].
SECTM1A is an immunoglobulin-like receptor whereas GITR is a member of the TNFR superfamily and is known to bind GITRL, a member of the TNF superfamily ligand group. All other TNFR superfamily members bind to ligands of the TNF superfamily. The only known exception to this rule is that of Herpes virus entry mediator (HVEM, TNFRSF14) which was shown to bind to B and T Lymphocyte Attenuator (BTLA, CD272) [10,11] an immunoglobulin superfamily member, on T cells, in addition to four other ligands, LIGHT, CD160, lymphotoxina and Herpes simplex virus glycoprotein D [12]. The binding of HVEM to BTLA on T cells is inhibitory to T cell activation [10].
TNF family ligands form non-covalent homotrimers which induce the receptor to adopt a trimeric conformation which enhances signalling [13]. In the case of the BTLA-HVEM interaction it has been shown that a 1:1 stoichiometry can exist between extracellular domains of BTLA and HVEM however BTLA exists as oligomers on the cell surface which may be capable of clustering HVEM to enhance signalling [14,15]. Human GITRL has also been shown to adopt a homotrimeric conformation [16] inducing homotrimeric clusters of GITR on the cell surface. Mouse and human GITRL differ in their propensity to form trimers in solution with mGITRL forming stable dimers [17]. The stoichiometry of the interaction between GITR and SECTM1A seen in our studies is currently unknown. Despite our demonstration that both SECTM1A-Fc and GITR-Fc fusion proteins are active in biochemical and cell binding experiments we did not detect a specific interaction between them by surface plasmon resonance.
Whether binding to CD7 or to GITR it is possible that 'reverse' signalling occurs within the cells expressing cell surface SECTM1A or SECTM1B, which include epithelial, endothelial and myeloid cells. Both proteins have short cytoplasmic tails containing tyrosine residues which do not conform to known signaling motifs. The possibility of reverse signalling remains to be excluded.
Functionally it has been reported that murine SECTM1B inhibits ConA-induced lymph node cell proliferation but not lymph node cell proliferation induced by anti-ab TCR antibodies [4]. We consistently observe inhibition of anti-CD3 induced T cell proliferation and IL-2 production with SECTM1B, and the converse with SECTM1A. These differences may be due to differences in cell preparation as we used highly purified negatively selected CD4+ cells compared to the previously published study.
The SECTM proteins join a large group of ligands used by tissues to co-stimulate or diminish T cell responses [18,19]. [1,20,21,22,23,24,25,26] [27]. Co-expression of the SECTM molecules in tissues may have alternate outcomes depending on the receptors expressed by different T cells. A degree of tissue regulation of SECTM1 has been shown by Lam and colleagues who demonstrated up-regulation of SECTM1 expression in response to IFNc by thymic epithelium [28].
In conclusion we have demonstrated opposite functional outcomes of SECTM1A and SECTM1B, widely expressed tissue ligands, binding to T cells. We have shown that SECTM1A represents a new CD7 ligand in mice in addition to binding to GITR and we raise the possibility that alternative receptors for both SECTM1A and SECTM1B exist on T cells.
CHO cells stably transfected with MHC class II (IE k ) were a kind gift from Prof Neil Barclay, Sir William Dunn School of Pathology Oxford.

Production of SECTM1 Fusion proteins
Fc fusion protein expression plasmids were constructed using an expression vector pEE12(CMV/T7)CD5Lhc1 encoding a CMV promoter and CD5 leader sequence directly 59 to the human IgG1 aglycosyl Fc coding region. cDNAs were cloned between the CD5 leader sequence and hIgG1 fragment.
Fusion proteins were produced in CHO-S cells and purified with protein G agarose with elution using glycine at pH 2.0.
Protein N glycosidase F digestions 20 mg of Fc fusion proteins were denatured and subjected to digestion with 1000 units of protein N glycosidase F (New England Biolabs) for 2 hours at 37uC according to the manufacturers instructions. Digested and un-digested fusion proteins were run on SDS PAGE, coomassie stained and western blotted for human Ig-Fc to assess their level of N-linked glycosylation.
Proliferation assays CD4+ T cells were stimulated with plate-bound anti-CD3 (145.2C11) and IL-2 or anti CD28 where indicated. SECTM1A, 1A-ExIV, SECTM1B-Fc fusion proteins or Campath-1H as an Fc control with identical human IgG1 aglycosyl Fc region were immobilised by air-drying onto 96 well culture plates as described by Stebbings et al [30]. 16105 CBA/Ca CD4+ T cells were used per well. Tritiated thymidine at 0.5 mCi per well was added sixteen hours before harvesting the cultures onto glass fibre filters and scintillation counting using standard techniques. ELISA IL-2 was measured by ELISA using a paired antibody set from BD Biosciences and recombinant IL-2 from R&D Systems to construct standard curves.

Flow cytometry
Fc fusion proteins used for FACS were labelled with allophycocyanin using a Lynx rapid conjugation kit (AbdSerotec) according to the manufacturer's instructions. Cells were incubated with fusion proteins at 10 ug/ml in PBS with 0.5% bovine serum albumin and 2 mM EDTA for 1 hour at 4C followed by three PBS washes and fixation in 2% paraformaldehyde.

Surface Plasmon Resonance Analysis
For surface plasmon resonance experiments a BIAcore 3000 instrument was used at 25uC. Rabbit anti-mouse antibody BR-1000-57 (BIAcore) was directly immobilized by amine coupling in 10 mM NaAc buffer, pH 5.5 onto CM5 chips. This was followed by binding of mouse anti human IgG (SBH2, a kind gift from Alison Tutt, Tenovus, Southampton) to all cells. CAMPATH-1H, mCD7-Fc, hIL7R-Fc, SECTM1A-Fc, SECTM1B-Fc, GITR-Fc and GITRL-Fc were bound to separate flow cells. Fc-fusion protein-containing solutions at 25 mM were injected in HEPESbuffered saline, pH 7.4, over the immobilized capture antibody-Fc fusion protein containing flow cells at a rate of 5 ml per minute.
Immunoprecipitation of SECTM1A interactors 5 mg of SECTM1A-Fc or hIL7R-Fc negative control fusion protein were incubated in PBS with 1 ml of protein G agarose (Invitrogen) for 1 hour at 4C. The agarose was then washed with PBS followed by 50 mM sodium borate and the fusion proteins cross-linked to the protein G with dimethyl pimelimidate at 7 mg/ ml in 0.2 M triethanolamine for 1 hour at room temperature. The columns were washed with 5 column volumes of 0.2 M triethanolamine followed by blocking with 5 column volumes of 0.1 M ethanolamine. Un-cross-linked fusion proteins were eluted from the column with 0.2 M glycine pH2.0. 2610 8 concanavilin-A-activated CD7 2/2 splenocytes were lysed on ice with 0.5% lauryl maltoside prior to clearing by centrifugation. Lysates were incubated with the columns for 2 hours at 4C followed by washing with 5 column volumes of PBS and elution with 0.2 M glycine. Eluates were concentrated with Centricon 10,000 molecular weight cutoff dialysis spin columns (Millipore) and run on SDS PAGE gels.

Mass Spectroscopy
SDS-PAGE gel slices were digested with trypsin and desalted on a C18 packed pipette tip. Samples were injected onto an Ultimate 3000 nano HPLC (Dionex) system coupled to an Orbitrap mass spectrometer (Thermo Electron). Samples were resolved on a 5 cm by 100 micron inner diameter picotip column (New Objective) which was packed in house with reprosil-Pur C18-AQ phase. A 40 minute gradient was used to separate the peptides. The mass spectrometer was operated in data dependent acquisition mode. Precursor scans were performed in the orbitrap at a resolving power of 60,000, from which five precursor ions were selected and fragmented in the linear ion trap. Charge state +1 ions were rejected. Peak lists were generated using DTAsupercharge and searched using Mascot (Matrixscience). Data were searched against MSDB database, restricting the taxonomy to mus musculus. Precursor mass accuracy tolerance was set at 10 ppm and MS/MS at 0.5 Da.