The PAIRS methodology aims at pairing cargos with their cargo receptors and to elaborate on the existing body of knowledge on cargo extraction from the ER. Our analysis did not include all possible cargos since the proteins in the GFP library are all C-terminally tagged and under their endogenous promoter. This resulted in a number of strains whose proteins were mislocalized due to the tag or had fluorescence levels below detection 
. This may explain why we could not identify some of the previously recognized cargo/cargo receptor pairs. Despite these caveats, the power of the PAIRS analysis is that it is not biased within the set of pre-selected proteins, allowing us a broad overview of cargo receptors function. This has allowed us to gain insight to the rules governing their specificity and to present the first step towards creation of a cellular “traffickome.”
We demonstrate the value of the PAIRS approach by identifying putative new cargo proteins for most of the known cargo receptors of yeast. This represents 31 of the 157 proteins tested, of which 27 had not been previously linked to a cargo receptor. This is likely to be a slight underestimate of the success rate as some of the proteins stated to have a vacuolar localization in the GFP Database probably reflect ER residents displaced by the tag and hence would not be expected to have a dedicated cargo receptor. One general trend seen across all hits is that removal of a receptor did not result in a complete block of ER exit of its GFP-tagged cargo. This is consistent with previous studies of known cargo/receptor pairs including GPI-anchored proteins 
. It may be that cargo receptors typically act to accelerate the exit of a particular cargo, and that some bulk flow always occurs, with the volume of this flow increasing for a particular protein if it accumulates in the ER.
The increased knowledge of the range of cargo for each specific cargo receptor should make it easier to generate hypotheses as to what determines the selective recognition of particular cargo by individual cargo receptors. Indeed, examination of the spectrum of cargo relying on the cargo receptor Erv14 suggested that this large and non-homogenous group is recognized on the basis of the length of its TMD. By assessing this hypothesis using one particular cargo, Mid2, we found that TMD length is a major determinant for allowing Erv14 to accelerate exit from the ER. It is still feasible that Erv14 recognizes a more specific motif in Mid2 adjacent to the TMD, and alterations in the length of the TMD affect the position of this region relative to the bilayer and thus prevent Erv14 binding. However, this seems unlikely given the wide range of bitopic and polytopic proteins that are affected when Erv14 is deleted, and their lack of shared sequence motifs. Interestingly, there have been two recent studies reporting that shortening the TMD of a mammalian protein reduces its exit rate from the ER 
. The mechanism for these effects was not determined, but one reporter used was VSV-G that has been found to depend on the Erv14 paralogue, CNIH4, for normal ER exit 
Whether Erv14 enables exit of long TMD-containing proteins from the ER by performing more than just COPII coupling is yet to be uncovered. One option is that it could also act as a chaperone to protect protruding hydrophobic residues on cargo proteins thus enabling them to assume a correct conformation in the shorter ER membranes. Another option is that it sorts long TMD containing proteins into areas of the ER that have thicker membranes thereby enabling their recruitment to vesicles. Regardless, it seems that speed of ER exit may be a major determinant in Erv14's function. In yeast, one of the substrates, Axl2, has been shown to require very rapid ER exit 
, as it must be inserted into the forming yeast bud at a particular point in the cell cycle (AXL2
mRNA is under cell-cycle control) 
. One of the substrates for the Drosophila
paralogue of Erv14, Cornichon, is Gurken, a TGFα-like bitopic protein 
. The ability of Gurken to polarize Drosophila
oocytes depends on its rapid exit from a restricted region of the oocyte ER following translation from a pool of mRNA that is spatially restricted for a short time during development 
. Indeed, also the action of Cornichon on Gurken requires that the latter has a TMD 
Our analysis of the substrate recognition mode for Erv14 reveals that TMD length-dependent sorting may be a more general principle in cellular trafficking than previously appreciated. Using our DxE-containing Mid2 variant we noticed that the TMD variants also underwent a TMD length-dependent sorting in the Golgi apparatus. This is consistent with a previous study examining the effects of lengthening the TMD of the yeast ER protein Ufe1, although this is the first time it has been shown to occur with a homogenous synthetic TMD rather than a native TMD, which may contain additional cryptic sorting motifs 
How might length-dependent sorting occur in the Golgi if it does not involve Erv14? One option is that a dedicated cargo receptor exists at this compartment that has not yet been identified. However, an alternative option is that the vesicle composition itself plays a major role in this step with lipids and/or cargo proteins directing a change in bilayer properties 
In summary, our unbiased approach allowed the formulation of a simple hypothesis for the underlying commonality allowing cargo identification by Erv14. Using Mid2 with a synthetic TMD has allowed us to indeed observe such TMD length-dependent steps both in the ER and the Golgi. The notion that TMD length is used by the cell to sort proteins is appealing 
, since many and diverse membrane proteins must be continuously extracted from the ER following synthesis. If these proteins share a generic feature that reflects their normal environment being different to that present in the ER, in this case TMD length, then it would provide a simple means of sorting of many different proteins without the need for specific linear signals.
More generally, the conceptual methodology that we have put forward here could be applied in a wider context to uncover protein localization changes that occur in the absence of any specific gene in the genome. The notion of the “traffickome” could be extended to other trafficking events such as retrograde Golgi to ER traffic, Golgi to plasma membrane traffic, or Golgi to vacuole traffic. Hence, by pairing high-throughput genetic manipulations with a microscopic output it is now possible to study basic questions of specificity and promiscuity in cell biology that have previously been difficult to tackle.