Identification of CD7 cis-interactors using the yeast split-ubiquitin membrane-based yeast two-hybrid system
Human CD7 has been shown to associate with CD45 and CD3 at the membrane surface (Lazarovits et al, 1994) and its cytoplasmic tail to bind PI3K and type II PI4K (Chan et al, 1997/Lee et al, 1996/Subrahmanyam et al, 2003). In each case, these interactions were discovered using co-immunoprecipitation with antibodies raised against hCD7. The lack of antibody against mouse CD7 precludes the use of this technique for identifying binding partners. Biochemical or biophysical methods used for the analysis of interactions between proteins include amongst others co-fractionation of protein complexes by chromatography, epitope tagging followed by mass spectrometry, fluorescence resonance energy transfer (FRET), surface plasmon resonance or X-ray crystallography (see Hooker et al, 2007 for review). Most of these techniques are time-consuming, expensive and are hardly suitable for large-scale studies. The yeast two-hybrid system (Y2HS) is a genetic approach that relies on the transcription of endogenous reporter genes of the yeast as a read-out for protein interactions (Fields & Song, 1989). Both halves of a transcription factor specific for the promoter of the reporter gene are brought together only when two proteins, each linked to one half of the transcription factor, are interacting. The reconstitution of a functional transcription factor allows the initiation of the transcription of the reporter gene which translates in growth on selective medium or enzymatic activity generating coloured compounds. This system is one of the most commonly used for the identification of novel protein interaction because it is easy to set up and it allows large-scale screenings of libraries. Y2HS accounts for more than 50% of all protein interactions described in the literature (Suter et al, 2006). Nevertheless, this technique has two major drawbacks originating from the same cause. One is that both events of protein binding and transcriptional output take place in the same subcellular compartment, the nucleus, leading to a high rate of false positives. The other one is that its use is mostly restricted to protein interactions occurring in the nucleus, therefore excluding of its scope the identification of interactions involving integral membrane proteins. The membrane-based split-ubiquitin yeast two-hybrid system (MbYTH) is a modified version of Y2HS that overcomes those issues. MbYTH take advantage of an ubiquitin-based split-protein sensor instead of the split-transcription factor used in the classical Y2HS (Johnsson and Varshavsky, 1994). Ubiquitin is a well conserved protein involved in the degradation of other proteins (Hershko,2005/Mayer, 2000). It serves as a tag allowing the proteasome machinery of the cell to recognise and degrade any protein to which it is covalently attached. Ubiquitin itself is not degraded in the process because endogenous ubiquitin specific proteases (UBPs) cleave the covalent link attaching the C-terminal end of ubiquitin to the protein being degraded, releasing the free ubiquitin to the cytosol for its recycling. Ubiquitin can be split in an N-terminal half (Nub) and a C-terminal half (Cub). Nub and Cub re-associate spontaneously to form a correctly folded and functional protein. By mutating an Ile to a Gly in position 3 in the wild-type Nub (NubI), the strong affinity between the two moieties is abolished. In the MbYTH system (Stagljar et al., 1998; Thaminy et al., 2003), the bait protein is expressed in frame with Cub and Cub itself is expressed in frame with LexA-VP16, an artificial transcription factor consisting of the DNA binding domain of LexA from E.Coli and the transactivator domain of VP16 from HSV. The mutated Nub (NubG) is expressed on the other hand in frame with the prey protein. If the bait and the prey proteins don't interact, there is no re-association of Cub and NubG and Cub remains in a misfolded conformation unrecognised by the UBPs. But as soon as the bait protein interacts with a prey protein, Cub and NubG are forced together to form a whole ubiquitin molecule which will be recognised by the UBPs, leading to the cleavage of the polypeptide chain between Cub and LexA-VP16. LexA-VP16 is then released and it translocates to the nucleus where it binds to the LexA operators lying upstream of a reporter gene. The VP16 transactivator domain recruits the RNA polymerase II complex to start the transcription of the reporter gene. Because the UBPs work physiologically in the cytosol, it makes the MbYTH system suitable for the study of integral membrane protein interactions (with either other membrane proteins or with cytosolic partners) as long as the Cub-LexA-VP16 part remains in the cytoplasm. Also, the system allows the compartmentalisation of the protein interactions and the transcriptional read-out. This is beneficial for the maintenance of a low background which is critical in large scale studies. We chose the Dualmembrane system from DualSystems Biotech (figure 3.1), which is the commercial version of the MbYTH system, to try to identify CD7 binding partners both at the membrane and in the cytosol. The host is NMY51, a strain of Saccharomyces Cerevisiae auxotrophic for L-leucine, L-tryptophan, L-histidine and L-adenine. LEU2 and TRP1, the genes encoding enzymes for the de novo synthesis of L-leucine and L-tryptophan (review in Pronk, 2002) are defective. Instead, HIS3 and ADE2 expression is under the control of the LexA operator, rendering their expression dependent on an interaction between bait and prey. The bait and the prey plasmids carry LEU2 and TRP1 respectively, allowing the complementation of the defective genes when transformed into NMY51. Transformants for which an interaction between bait and prey occurs can selectively be grown on media lacking L-leucine, L-tryptophan, L-histidine and/or L-adenine. Additionally, a bacterial LacZ gene, encoding for the b-Galactosidase, has been integrated in the URA3 locus under the control of the LexA operator and serves also as reporter gene.
4.2.1 Construction and cloning of mCD7 bait plasmid
Mouse CD7 was expressed as a C-terminal ubiquitin-LexA-VP16 fusion in the pBT3-STE vector (figure 3.2). The image clone 4985794 (Open Biosystems), which is the complete coding DNA sequence of mouse CD7 (mCD7) cloned in pCMV.Sport6 (Invitrogen), was used as a template for PCR amplification. The primers mCD7pBT3S_5 and mCD7pBT3C_3 (table 2.1) were designed to amplify CD7 lacking its signal sequence, the start codon and the stop codon, to introduce a Sfi I restriction site at both ends and to keep CD7 sequence in frame with the STE2 leader sequence as well as with the Cub-LexA-VP16. The PCR amplicon and the pBT3-STE vector were digested with Sfi I, purified and ligated before to be transformed into TOP10 E.Coli cells. There are two Sfi I sites in the multiple cloning site of the vector. The two Sfi I sites have different overhangs. The incorporation by PCR of the corresponding Sfi I sites at both ends of the mCD7 coding sequence allows its directional cloning into the vector. Several colonies resulting from the transformation were grown and their plasmid isolated. These plasmids were subjected to DNA sequencing and one clone with the correct sequence was used to transform the yeast strain NMY51. The resulting transformant was called NMY51:pBT3-STE/mCD7 and kept under selective pressure by growing in minimal base SD media lacking L-leucine (SD-leu).
4.2.2 Verification of the correct topology of mCD7 expressed in NMY51
The correct expression and targeting to the membrane of CD7 was assessed by western blot (figure 3.3) and by using a functional assay. In the latter, the strain NMY51 was co-transformed with pBT3-STE/mCD7 or the control bait plasmid pCCW-Alg5 along with the control prey plasmids pAI-Alg5 or pDL2-Alg5.
Alg5 (asparagine-linked glygosylation 5) is a yeast protein member of the glycosyltransferase 2 family, resident of the endoplasmic reticulum membrane. pCCW-Alg5 expresses Alg5 as a fusion with Cub-LexA-VP16 while pAI-Alg5 and pDL2-Alg5 express Alg5 as a fusion with NubI and NubG respectively. Therefore, the co-expression of pAI-Alg5 and a protein properly inserted in the membrane, with the Cub-LexA-VP16 moiety located in the cytosol, should activate the reporter genes due to the strong affinity of NubI for Cub. On the other hand, co-expression with pDL2-Alg5 shouldn't, as Alg5 doesn't bind to itself and is unlikely to bind to a mammalian protein. Indeed, as shown in figure 3.4, NMY51 could grow on SD-leu/-trp/-his/-ade agar media when pBT3-STE/mCD7 or pCCW-Alg5 were co-transformed with pAI-Alg5, but not when they were co-transformed with pDL2-Alg5.
4.2.3 Library screening and identification of putative interactors
Two split-ubiquitin libraries were constructed by Dualsystems Biotech AG (Zurich, Switzerland). The starting material was total RNA extracted from T cells of female A1.RAG mice (cf. table 2.2) cultured in the presence of mytomicin C-treated bone-marrow derived dendritic cells (BMDCs), exogenous human recombinant TGF-b and DBY, a peptide of the male antigen H-Y. After seven days, the T cells were collected and isolated from the BMDCs by Ficoll centrifugation and re-cultured in IL-2 at 2000 U/ml for a further week. Such population of cells was termed DBYT.IL2 (ref Nolan et al,?, more detailed explanation in intro?). The cDNA of DBYT.IL2 was directionally cloned into pPR3-C and pPR3-N to create the libraries DBYT.IL2/X-NubG and DBYT/NubG-X respectively (Howie at al, 2009). NubG-X libraries are used to identify cytosolic and type II integral membrane proteins while X-NubG libraries are suitable for identifying cytosolic and type I integral membrane proteins. Both libraries were interrogated for interactions with CD7. One issue inherent to genetic-based methods to assess protein interactions is the high rate of false positives resulting from the activation of the reporter genes in the absence of bait-prey interactions. Such non-specific activation can be due to the prey proteins as well as the bait protein. The self-activation of the bait protein may rise from its overexpression, its lability or its cleavage by endogenous proteases. We assessed the self-activation of CD7-Cub-LexA-VP16 by conducting a pilot screen with the vectors pPR3-N, pPR3-C and pDL2-Alg5 as negative control preys, using increasing concentrations of 3-AT in the selective media agar plates. The empty prey vectors pPR3-N and pPR3-C encode a non-sense peptide fused to NubG either at the N-terminal or C-terminal end of the peptide. Plates showing a strong growth or no growth at all were discarded as not stringent enough or too stringent respectively. Plates with 2 to 4 colonies were considered as having the degree of stringency necessary to keep a low background of growth while still allowing the detection of weak true interactions. The screening of the libraries was therefore carried out on SD-Leu/-Trp/-His/-Ade agar plates complemented with 10mM 3-AT. Despite this, a high-throughput sequencing methodology revealed that at least 30% of the clones were false positives. The most common were clones expressing ubiquitin, b-2 microglobulin and ribosomal or mitochondrial proteins (for a review of common false positives see http://www.fccc.edu/research/labs/golemis/InteractionTrapInWork.html). Two hundred clones for each library were sequenced. A list of putative interactors that came out of the screenings and the number of hits they generated is shown in table 3.1. We investigated further a shortlist of putative interactors by confronting them to a bait-depencency test. In this test, the prey vectors isolated from the screening transformants were re-transformed into the mCD7-bearing yeast and tested for interaction on selective agar plates. As negative controls, the prey vectors pPR3-N, pPR3-C and pDL2-Alg5 were also transformed. The pAI-Alg5 prey vector was used as positive control for transactivation of the reporter genes. Finally, the bait vector pTSU2-APP and the prey vector pNubG-Fe65 were used as positive controls for protein interaction. Indeed, APP (amyloid beta A4 precursor protein) has been shown to bind to the cytosolic protein Fe65 (amyloid beta A4 precursor protein-binding family B member 1) (Matsuda et al, 2005) and their interaction within the yeast reporter strain results in a robust growth on selective plates. The clones that were able to grow on highly stringent selective agar plates were considered as true interactors and are listed in table 3.2. Interactions of these candidates with CD7 will need to be investigated in mammalian cells using a classical co-immunoprecipitation approach for those candidates whose commercial antibodies are readily available.