Analysis of endoplasmic reticulum trafficking signals by combinatorial screening in mammalian cells

Abstract

To improve the accuracy of predicting membrane protein sorting signals, we developed a general methodology for defining trafficking signal consensus sequences in the environment of the living cell. Our approach uses retroviral gene transfer to create combinatorial expression libraries of trafficking signal variants in mammalian cells, flow cytometry to sort cells based on trafficking phenotype, and quantitative trafficking assays to measure the efficacy of individual signals. Using this strategy to analyze arginine- and lysine-based endoplasmic reticulum localization signals, we demonstrate that small changes in the local sequence context dramatically alter signal strength, generating a broad spectrum of trafficking phenotypes. Finally, using sequences from our screen, we found that the potency of di-lysine, but not di-arginine, mediated endoplasmic reticulum localization was correlated with the strength of interaction with alpha-COP.

Figure 1

Strategy for screening combinatorial libraries of sorting signal variants in intact mammalian cells. (A) Schematic depiction of the CD4-GFP reporter construct. GFP fluorescence is used to assess total reporter protein expression for individual cells, a phycoerythrin (PE) conjugated anti-CD4 antibody is used to measure CD4-GFP surface levels. Stable HEK293 cell lines expressing the reporter with an ER retention/retrieval sequence or an inactive version of the signal were created by Flp recombinase-mediated integration into the same genomic locus. Cells from the two lines were mixed, stained with the anti-CD4-PE antibody, and analyzed by flow cytometry (1,000 cells shown). Two populations reflecting the two trafficking phenotypes can readily be distinguished. (B) Flow chart detailing the individual steps of the screening strategy.

Figure 2

Libraries of combinatorial variants of di-lysine and arginine-based ER retention/retrieval signals. Schematic depiction of the three libraries analyzed, the randomized positions are indicated in the C-terminal tail. The frequencies of the individual bases were determined by sequencing six to eight randomly picked clones from the unselected libraries. Populations of 3T3 cells infected with each library were sorted on a FACS Vantage SE cell sorter. Gates R2 and R3 were chosen to separate cells according to the two trafficking phenotypes of the reporter protein (retained or expressed on the surface). The plots shown are gated on viable cells as determined by propidium iodide staining (gate R1, not shown).

Figure 3

Amino acid frequency at flanking positions of di-lysine and arginine-based signals. (A) Frequency plots for each randomized position in the di-lysine signal libraries. Unique sequences identified for both libraries were analyzed with respect to the amino acid frequency at each randomized position in the pool of intracellularly retained (x axis) or surface-expressed (y axis) reporter proteins. The number of unique sequences were 56 (-KKXX-C, retained), 48 (-KKXX-C, surface), 39 (-XXXKTN-C, retained) and 39 (-XXXKTN, surface). The position analyzed in the respective plot is indicated in green. Amino acids favored in the surface pool are indicated in red, whereas amino acids favored in the retained pool are shown in blue. Amino acids shown in black were either rare or equally abundant in both pools. (B) Table of clones selected from the library of arginine-based signal variants. Blue randomized positions indicated ER localization, whereas red randomized positions indicate surface expression. Tables 1 and 2 (which are published as supplementary material on the PNAS web site, www.pnas.org) list the original sequences obtained from the screen.

Figure 4

Quantitative assays for TGN processing and surface expression of selected library clones. Fixed positions are shown in black, sequences of tails conferring ER retention/retrieval are shown in blue, and sequences of tails allowing for surface expression are shown in red. TGN processing was assessed by measuring alkaline phosphatase activity in the supernatant of transiently transfected COS-7 cells (left column). Surface expression was quantified by antibody staining and luminometry (right column).

Figure 5

Cellular staining patterns of selected signal variants. (A) Costaining of selected clones with antibodies directed against marker proteins for the ER (anti-SRP54/anti-Grp75), Golgi (anti-p115), and the TGN (anti-γ-adaptin). Fluorescence of CD4-GFP fusion proteins is merged with red fluorescence of stained marker proteins (Cy3-conjugated secondary antibodies). (B) GFP staining patterns for di-lysine signals with only the -1 or -2 position varied and tails containing an RKR signal with different flanking residues. Each series covers the whole range from mainly ER to surface localization. (Bar = 25 μm.)

Figure 6

Yeast two-hybrid analysis of interaction between selected tails from the screen and subunits of the COPI vesicle coat. By using an active and inactive version of the di-lysine or the arginine-containing tail as a bait, all seven subunits of the COPI vesicle coat were tested for interaction (A and B). Indicated constructs were cotransformed and colonies were streaked on selective medium (dropout without histidine and adenine) after 3 days. Pictures were taken after 4 days of growth on selective medium. Specific interaction with the α-COP subunit was observed for the -KKYL-COOH (not the -KK-COOH) tail (A), but not for the arginine-containing tail (B). Bait constructs containing different versions of the di-lysine signal were tested against the α-COP subunit (C). Strong versions of the signal isolated from the screen interacted with α-COP, whereas signals of intermediate strength or inactive versions gave no signal.