Peptide Signals Encode Protein Localization

Many bacterial proteins are localized to precise intracellularlocations, but in most cases the mechanism for encoding localizationinformation is not known. Screening libraries of peptides fusedto green fluorescent protein identified sequences that directedthe protein to helical structures or to midcell. These peptidesindicate that protein localization can be encoded in 20-amino-acidpeptides instead of complex protein-protein interactions andraise the possibility that the location of a protein withinthe cell could be predicted from bioinformatic data.

INTRODUCTION

Top

INTRODUCTION

Many proteins have precise intracellular addresses in bacteria,i.e., they are restricted to an area measured in tens of nanometersand not merely to domains such as the cytoplasm or membrane(19, 27, 30). Bacterial protein localization can change in responseto external stimuli or internal genetic programs, and essentialprocesses such as cell division and differentiation depend oncorrect protein localization (13, 23). It is clear that structuralproteins must be localized to form functional assemblies, suchas flagella, pili, and the cytokinetic Z ring. Yet, many regulatoryproteins have a specific intracellular address as well. Studiesof several bacterial species, including Caulobacter crescentus,have revealed the localization of histidine kinases and transcriptionfactors, and in many cases localization is required for properfunction of these proteins in vivo (19). In addition, the identificationof bacterial analogs of tubulin (12, 26), actin (20, 36), andintermediate filaments (2) indicates that bacteria have a cytoskeletonand suggests that there may be a high degree of intracellularorganization (32).

A priori, there are several possible ways that protein localizationcould be specified. Each protein could be localized througha unique set of interactions with other localized proteins.This mechanism has been observed for some chemoreceptors (22),proteins associated with the FtsZ ring at the cell divisionplane (13), and a histidine kinase (28). Alternatively, proteinlocalization could be mediated by global sorting pathways inwhich localized proteins contain sequence or structural informationthat targets them to a particular site through interactionswith general sorting machinery or cytoskeletal filaments. Thistype of mechanism is used for nuclear localization in eukaryotes(18, 25) and for protein secretion in bacteria (5). Any globalmechanism would require homologous sequence or structural features,but protein localization signals have not been reported.

One simple mechanism to direct protein fate is through peptidesignal sequences. The N-end rule and tmRNA peptide signals targetproteins for degradation (21, 35), and N-terminal signal sequencestarget proteins to the Sec or Tat (twin-arginine translocation)pathway for secretion (6, 10). In each of these examples, thepathways were uncovered by genetic studies, first to identifythe cis information that targets the substrate to the pathwayand then to identify the trans-acting components of the cellularexport machinery. In principle, peptide-encoded informationcould also target proteins to a specific location. Here, wetest whether short peptides contain sufficient information tolocalize proteins and find that multiple complex protein localizationpatterns can be produced by 20-amino-acid peptides.

MATERIALS AND METHODS

Top

MATERIALS AND METHODS

Strains and plasmids. The C. crescentus strain used in this work was CB15N (14). Aplasmid library was constructed by amplifying the egfp gene(Clontech) by PCR using the primers GGAATTCCATATG(N)₆₀GTGAGCAAGGGCGAGGAGCTGTTCand CTAGTTCTAGATTACTTGTACAGC and cloning the product into pJS14(33) under control of a xylose-inducible promoter (24) to generateN₆₀-egfp. Escherichia coli cells were transformed with the plasmidlibrary and grown as described previously (9), resulting in $~$ 2 x 10⁵ colonies. E. coli colonies were pooled, and plasmidDNA was prepared and mobilized into C. crescentus by electroporation.To generate LP2-mCherry, the mCherry coding sequence was clonedinto pJS14, and a DNA fragment containing the xylose-induciblepromoter and the LP2 sequence was amplified by PCR and clonedinto a BamHI site at the 5' end of mCherry. Due to the BamHIsite, the LP2-mCherry protein contains a glycine-serine linkerbetween LP2 and mCherry.

Screen for localization signals. Cells from the C. crescentus library were grown to an opticaldensity at 660 nm of 0.30 to 0.40 in PYE medium (11) with 0.3%xylose to induce expression of N₆₀-egfp and visualized usingepifluorescence microscopy. Localization patterns were confirmedby growing independent cultures in PYE with xylose, or in M2X(24), and examining the cells with epifluorescence microscopy.In all cases, the localization patterns were unchanged in differentmedia. Plasmid DNA was prepared from cells containing localizedfluorescence signal and sequenced to identify the N-terminalpeptide fused to green fluorescent protein (GFP). The effectof A22 was assayed by adding the inhibitor to a final concentrationof 10 µg/ml for 1 hour prior to observation.

Microscopy. Cells were immobilized on 1% agar and imaged using an EclipseE600 microscope (Nikon) with a 100x Plan Fluor oil numericalaperture 1.30 objective in conjunction with a CoolSNAP fx charge-coupleddevice camera (Photometrics) controlled by Image-Pro Discoverysoftware (Media Cybernetics). Images were deconvolved usingthe 2D Blind Deconvolution algorithm in AutoQuant software (MediaCybernetics).

For optical sectioning, an IX70 laser scanning confocal microscopewith a 100x UPlanFL oil numerical aperture 1.3 objective andFluoview software (Olympus) was used to obtain a Z-series ofconfocal images at increments of 0.15 µM. Each seriesof optical sections was deconvolved through 30 iterations usingAutoQuant software.

RESULTS AND DISCUSSION

Top

RESULTS AND DISCUSSION

Identification of peptide localization signals. To screen for peptides that contain a localization signal, alibrary of genes containing 20 redundant codons fused to the5' end of a gene encoding GFP was expressed in C. crescentus,and the cells were screened for localized GFP fluorescence.Cells producing GFP without an N-terminal peptide extensionhad fluorescence distributed evenly throughout the cell (Fig.1A), as did the vast majority of clones from the library, demonstratingthat GFP has no inherent localization signals. Five of the $~$ 3,000clones had specific patterns of fluorescence within the cells,and three of these were investigated in more detail.

View larger version (28K):
[in this window]
[in a new window]

FIG. 1. Peptides can specify discrete locations within the bacterial cell. C. crescentus cells producing GFP (A) or GFP with a 20-amino-acid peptide at the N terminus (B to D) were imaged by epifluorescence (left panels) or differential interference contrast (right panels) microscopy. The position of the focus of fluorescence in cells producing LP3-GFP, shown by schematic diagram (D, bottom left), was measured as a fraction of the distance from the stalked pole (fractional distance = 0) to the flagellar pole (fractional distance = 1.0), and a histogram is shown (D, bottom right).

Epifluorescence images of cells producing LP1-GFP showed a patternof three to four bands extending across the short axis of thecell (Fig. 1B). Similar patterns have been observed in fluorescenceimages by proteins localized in a three-dimensional helicalstructure (1, 8, 16, 31, 32, 34). To determine if LP1-GFP waslocalized in a three-dimensional structure, optical sectionswere obtained by taking confocal images at 0.15-µm incrementsalong the z axis (Fig. 2). These optical sections revealed twohelices of fluorescence spanning the length of the cell.

View larger version (19K):
[in this window]
[in a new window]

FIG. 2. LP1-GFP localizes to two helices in C. crescentus. Confocal microscopy was used to acquire a Z-series of fluorescence images of C. crescentus producing LP1-GFP. Schematic diagrams indicate the relative position of the optical section. A differential interference contrast image (top) shows the cell boundaries.

The epifluorescence images of cells producing LP2-GFP showedan array of spots (Fig. 1C), a pattern which is also consistentwith a three-dimensional structure (1, 8, 16, 31, 32, 34). Opticalsections of these cells revealed a helix of GFP fluorescencealong the long axis of the cell (Fig. 3). The pattern producedby LP2-GFP is clearly distinct from the pattern produced byLP1-GFP, demonstrating that 20-amino-acid peptides target GFPto at least two different helical structures in the cell.

View larger version (19K):
[in this window]
[in a new window]

FIG. 3. LP2-GFP localizes to a helix in C. crescentus. Confocal microscopy was used to acquire a Z-series of fluorescence images of C. crescentus producing LP2-GFP. Schematic diagrams indicate the relative position of the optical section. A differential interference contrast image (top) shows the cell boundaries.

A third fusion protein, LP3-GFP, produced a single focus offluorescence signal near the middle of each cell (Fig. 1D).To more precisely determine the position of the LP3-GFP focus,the distance from the stalked pole was measured in 54 cells.In 87% of the cells the focus was one-fourth to one-half ofthe distance from the stalked pole to the flagellar pole (Fig.1D). Similar patterns have been observed for the chromosomalreplication machinery and proteins required for cytokinesisand chromosome segregation (27, 32), although it is not clearthat LP3-GFP colocalizes with any of these structures.

Helices do not require MreB filaments. Helical patterns have been observed for the actin-like proteinMreB in several species of bacteria, including C. crescentus(16, 32). When LP1-GFP or LP2-GFP was produced in cells thatcontained MreB fused to the red fluorescent protein mCherry,the GFP and mCherry fluorescence signals were not colocalized(data not shown), indicating that the LP1 and LP2 peptides werenot directing GFP to the MreB helix. The MreB filaments canbe disrupted in vivo by the addition of the small molecule A22(17). When A22 was added to cells producing LP1-GFP or LP2-GFP,the GFP helices remained intact for more than 1 hour (Fig. 4).Under the same conditions, the mCherry-MreB signal was completelydispersed. These results confirm that localization mediatedby LP1 and LP2 does not require MreB filaments. Electron tomographystudies of C. crescentus showed several MreB-independent filaments,although the composition of these filaments has not been determined(7). LP1 and LP2 may be targeting GFP to one of these filamentsor to a previously unobserved structure.

View larger version (37K):
[in this window]
[in a new window]

FIG. 4. LP peptide-directed protein localization is not affected by the MreB inhibitor A22. C. crescentus cells producing mCherry-MreB (A) or GFP with a localizing peptide at the N terminus (B and C) were imaged by epifluorescence (left panels) and differential interference contrast (right panels) microscopy before (–A22) and after (+A22) treatment with A22 for 1 h.

Localization is independent of GFP. It is likely that GFP does not contribute to the localizationsignal in the LP fusion proteins, because GFP with no N-terminalpeptide has no localization and the vast majority of 20-merpeptides do not induce localization, whereas the three LP peptideseach produce a different pattern. To determine if other proteinscan also be localized by LP signals, LP2 was fused to mCherry,which has little sequence similarity to GFP (29). Localizationof LP2-mCherry was indistinguishable from that of LP2-GFP (Fig.5), demonstrating that the information contained in the LP2sequence is sufficient for localization of two distinct proteins.

View larger version (65K):
[in this window]
[in a new window]

FIG. 5. LP2 localizes mCherry to the same pattern as GFP. LP2-GFP and LP2-mCherry were imaged by epifluorescence (left panels) and differential interference contrast (right panels) microscopy. The localization patterns are indistinguishable, indicating that LP2 is sufficient to localize two different proteins to the same pattern in C. cresentus.

Protein localization produced by locons. The precise localization patterns produced by the LP peptidesdemonstrate that protein localization can be encoded in a shortamino acid sequence. We suggest naming these signals "locons,"by analogy with the "degron" degradation signals (37). Althoughthe possibility that the LP1 and LP2 peptides polymerize intofilaments has not been excluded, the small size of these peptideslimits the surface area available for protein-protein interactionsand makes polymerization improbable. The sequences of the LPpeptides suggest that they do not interact with known proteinsecretion systems. The signal sequence prediction programs PSORT(15), SignalP (3), and TatP (4) all predict that each of theLP-GFP fusion proteins is cytoplasmic. It is more likely thatthe localizing peptides bind to unknown structures in the cellor interact with an uncharacterized protein sorting system.No C. crescentus protein contains an exact match to one of thecomplete LP peptides, but many proteins have seven to nine residuesin common with one of the peptides, raising the possibilitythat localization signals for endogenous proteins are containedin a shorter peptide sequence or are partially redundant. Thediscovery and further characterization of the LP localizationsignals will permit the placement of biophysical probes andother molecules at particular sites within the cell and allowprotein localization to be predicted from bioinformatic data.

ACKNOWLEDGMENTS

Thanks to Elisabeth Mahen for help with strain construction;to Elaine Kunze and Nicole Bem at the Center for QuantitativeCell Analysis at the Huck Institutes of the Life Sciences, PennState University, for assistance with confocal microscopy andvisualization software; to Zemer Gitai for providing the mCherry-GFPstrain; and to Patrick Viollier for providing A22.

This work was supported by NIH grant GM68720.

FOOTNOTES

* Corresponding author. Mailing address: 401 Althouse, University Park, PA 16827. Phone: (814) 863-0787. Fax: (814) 863-7024. E-mail: kkeiler{at}psu.edu

Published ahead of print on 31 August 2007.

REFERENCES

Top