Fluorescent protein-based technologies: shedding new light on the plant endomembrane system


  • Imogen Sparkes,

    1. Biosciences,College of Life and Environmental Sciences, Geoffrey Pope, University of Exeter, Stocker Road, Exeter EX4 4QD, UK
    2. Department of Biological and Medical Sciences, Oxford Brookes University, Gipsy Lane, Oxford OX3 0BP, UK
    Search for more papers by this author
  • Federica Brandizzi

    Corresponding author
    1. Michigan State University–DOE Plant Research Laboratory and Department of Plant Biology, Michigan State University, East Lansing, MI 48824, USA
      (fax +517 353 9168; e-mail fb@msu.edu).
    Search for more papers by this author

(fax +517 353 9168; e-mail fb@msu.edu).


Without doubt, GFP and spectral derivatives have revolutionized the way biologists approach their journey toward the discovery of how plant cells function. It is fascinating that in its early days GFP was used merely for localization studies, but as time progressed researchers successfully explored new avenues to push the power of GFP technology to reach new and exciting research frontiers. This has had a profound impact on the way we can now study complex and dynamic systems such as plant endomembranes. Here we briefly describe some of the approaches where GFP has revolutionized in vivo studies of protein distribution and dynamics and focus on two emerging approaches for the application of GFP technology in plant endomembranes, namely optical tweezers and forward genetics approaches, which are based either on the light or on genetic manipulation of secretory organelles to gain insights on the factors that control their activities and integrity.


The secretory pathway comprises a number of morphologically and functionally distinct organelles, including the endoplasmic reticulum (ER), Golgi apparatus, trans-Golgi network, endosomes, vacuoles and plasma membrane, which work together to produce essential cellular components and to communicate with the external environment (Matheson et al., 2006). The endocytic routes allow cells to counterbalance the forward flow of membranes and proteins in the secretory pathway and to internalize external cues (Perez-Gomez and Moore, 2007).

Challenges faced in studying such a dynamic system in plants have required the development of tools which can be stably incorporated into the membrane or lumen of the compartment under observation. Prior to the development of genetically encoded fluorescent probes, dyes such as FM4-64 and DiOC6 were routinely used to stain membranes in live cells. However, a limitation of these lipophilic live-cell stains is that they can incorporate in a variety of membranes in a time-dependent manner, making it very difficult to study specific organelles. Also, owing to the rigidity and low porosity of the plant cell wall and the autofluorescent nature of vegetative tissue, several dyes that have been developed for use in other model systems are not transferable into plants. In addition, highly vacuolated cells are difficult to fix, making ultrastructural studies and histological analyses very problematic. While plant cell architecture offers inherent challenges compared with other model systems, it has several advantages for studying the endomembrane system: plant epidermal cells are large, and huge vacuoles exclude the cytoplasmic contents to the periphery. Based on the limitations of optical resolution, both of these points allow sizeable areas of the endomembrane system to be imaged in each cell with relative ease. Here we review some of the most advanced techniques and findings based on fluorescent protein technology for the study of the organization and dynamics of plant endomembranes. We will focus on optical tweezers and GFP-based forward genetics screens, but will also provide a brief overview of exciting findings obtained using other approaches such as fluorescence (or Förster) resonance energy transfer (FRET), fluorescence recovery after photobleaching (FRAP), photoactivatation and new fluorescent proteins which have also been discussed in a recent review paper (Sparkes et al., 2011). Technical challenges face plant scientists, such as overcoming the diffraction limit of light through the use of super-resolution microscopes (Fitzgibbon et al., 2010; Gutierrez et al., 2010), and enhancing signal to noise ratios through the use of variable angle epifluorescence microscopy (VAEM) or true total internal reflection fluorescence microscopy (TIRFM; Konopka and Bednarek, 2008; Staiger et al., 2010; Vizcay-Barrena et al., 2011). We expect that these technologies will be soon implemented in the study of the plant endomembranes.

Protein quantification, co-localization and dynamics

With a large availability of fluorescent proteins and publicly available imaging software (e.g. ImageJ, http://rsbweb.nih.gov/ij/), it is possible to perform not only co-localization analyses but also quantitative co-localization measurements, which can lead to unexpected results such as the identification of a novel compartment, the late pre-vacuole (Foresti et al., 2010). Other strategies have also enabled the quantification of protein expression by using a system based on a 20-residue self-cleaving 2A peptide from Foot and Mouth Disease Virus (Samalova et al., 2006). In this system, it is possible to generate polyprotein constructs in which fluorescent protein fusions to a reference marker and to the protein of interest are initially joined by the 2A peptide and then separated upon self-cleavage of such a peptide. The simultaneous visualization of the two fluorescent fusions allow ratiometric assays to quantify expression of the protein of interest, which is present in a cell in stoichiometric quantity with respect to the reference marker (Samalova et al., 2006).

Protein dynamics can also be monitored by FRAP. Here, fluorescent molecules are irreversibly bleached by a laser and the exchange into the region of bleached fluorescent molecules with unbleached variants is monitored over time to give an indication of protein movement (Lippincott-Schwartz et al., 2003). Similarly, movement can be monitored by using photoactivatable GFP (PAGFP) which has an increased level of fluorescence upon activation with an ultraviolet (UV) laser (Patterson and Lippincott-Schwartz, 2002). Unlike FRAP, the movement of activated PAGFP is monitored from the activation spot rather than into the bleached area. The types of movement under study could represent lateral motion in the membrane or through import/integration of ‘new’ fluorescent fusion molecules into the bleached area. For example, trafficking of a Golgi marker (ST-GFP) from the ER was assessed through FRAP and shown to be energy dependent, but independent of actin or microtubules (Brandizzi et al., 2002). Lateral movement of an ER membrane marker within the ER was shown to be independent of cytoskeletal elements (actin and the tail domain of the class XI myosin XIK) using both FRAP and photoactivation techniques (Runions et al., 2006; Sparkes et al., 2009a). Lateral mobility of AtFormin 1 in the plasma membrane was shown to be dependent on interactions with the cell wall (Martinière et al., 2011a). Similarly, such studies can quantify rates of diffusion and the local environmental factors that affect it. For example, the diffusion rate of a plasma membrane marker was affected by membrane viscosity and temperature (Martinière et al., 2011b). Also, mutant variants of the plant nuclear membrane markers AtSUN1 and AtSUN2 highlighted the importance of the amino terminal region and coiled coil regions in their lateral mobilities (Graumann et al., 2010).

New classes of fluorophores that are photoswitchable or photoconvertible have enabled users to study the movement of specific organelles within a group and protein populations, akin to biochemical pulse chase. For example, Kaede is a photoconvertible protein whose emission changes from green fluorescence to red upon irreversible conversion with UV light. Using these basic characteristics, it has been possible to monitor the dynamics of Golgi membrane turnover and incorporation into the phragmoplast (Brown et al., 2010). PIN3 trafficking at the plasma membrane has been reported through the use of the photoconvertible protein EoS (Kleine-Vehne et al., 2010). Dronpa is a photoswitchable fluorophore that is activated with a UV laser and switched off with a 488 nm laser. Using Dronpa, the dynamics of Secretory Membrane Carrier Protein 2 (SCAMP2) in the cell plate and plasma membrane of cultured tobacco cells was assessed (Toyooka and Matsuoka, 2009).

Green fluorescent protein as an in vivo alternative to biochemistry to study protein–protein interaction and protein topology in vivo

Green fluorescent protein technology has enabled the study of protein–protein interactions in vivo (Figure 1), overcoming some of the limitations linked to protein extraction and manipulation in biochemical assays. For example, it is now possible to monitor protein–protein interactions through FRET (Wallrabe and Periasamy, 2005). The FRET technique requires that the two proteins under study are fused to a donor or an acceptor fluorophore, where the donor’s emission spectrum overlaps with the excitation spectrum of the acceptor. Common fluorophore pairs are GFP with monomeric red fluorescent protein (mRFP), and cyan fluorescent protein (CFP) with yellow fluorescent protein (YFP). Upon interaction, the energy from the donor is non-radiatively transferred to the acceptor, resulting in acceptor fluorescence. For this to occur, the fluorophore pairs have to be within 1–10 nm of each other. There are several ways in which FRET can be monitored; measuring the subsequent fluorescence of the acceptor upon excitation of the donor, or monitoring the increase in fluorescence of the donor upon photobleaching of the acceptor (acceptor photobleaching). The FRET method can be prone to issues relating to relative concentrations of the protein under study and spectral bleed-through. To overcome these pitfalls, FRET can be monitored by fluorescence lifetime imaging microscopy (FLIM) (Wallrabe and Periasamy, 2005). Upon excitation, the donor fluorophore maintains its excited state for a defined period of time before returning to the ground state; this represents its unique lifetime. The lifetime of the donor is reduced upon interaction with the protein fusion, bringing the donor and acceptor fluorophores within 1–10 nm of one another. This technique has been used to monitor interaction of potential Golgi tethering factors (Golgins) with small GTPases, and to observe dimerization between proteins that shape ER membranes (reticulons) (Osterrieder et al., 2009; Sparkes et al., 2010).

Figure 1.

 Schematic representation of the fluorescent protein technologies used to assess protein topology and interactions.
Interaction between protein 1 and protein 2 monitored through fluorescence (or Förster) resonance energy transfer (FRET) (a) and biomolecular fluorescence complementation (BiFC) (b) are depicted.
(a) In FRET, upon interaction, energy from the excitation of the donor (blue star) is transmitted non-radiatively to the acceptor (yellow star) resulting in acceptor fluorescence. Monitoring of FRET can be undertaken in several other ways including quantifying the level of donor fluorescence upon acceptor photobleaching.
(b) In BiFC the two halves of a fluorochrome (yellow half stars) are fused to the two proteins of interest. Upon interaction between proteins 1 and 2 the two halves of the fluorochromes are no longer spatially separated, resulting in fluorescence upon excitation (whole yellow star).
(c) Using the same basic concept of spatial separation of the two halves of the fluorochrome, BiFC can be used to ascertain membrane protein topology.
(d) The application of redox-sensitive GFP (roGFP; green star) to monitor protein topology is highlighted for a membrane protein resident in the endoplasmic reticulum Ratiometric imaging of the emission of roGFP after sequential excitation at 405 and 488 nm can be used to determine whether the roGFP is in a reducing or oxidizing environment, and hence the topology of the protein to which it is fused can be determined.

A complementary technique to study whether proteins can form interacting pairs is biomolecular fluorescence complementation (BiFC), which is based on the expression of complementary halves of split fluorescent proteins fused to target proteins (Figure 1). If the latter form interacting pairs, the fluorescence signal would appear (Walter et al., 2004; Ohad et al., 2007). This approach has been successfully expanded to establish membrane protein topology (Zamyatnin et al., 2006). Since the fluorescent protein halves can associate in a crowded environment independently from the interaction of the proteins to which they are fused, BiFC can be used to establish the position of the protein to which the fluorescent proteins are fused with respect to a membrane. For example, half a YFP can be expressed in a cell along with a membrane protein fused to the other half of the YFP. If the latter is exposed to the cytosol, reconstitution of the fluorescence is indicative of the exposure of the protein terminus, to which the half YFP is fused, to the cytosol. This approach has led, for example, to the finding that an ER-associated putative membrane-anchored transcription factor of the MYB family is oriented with the TF domain exposed to the cytosol (Slabaugh et al., 2011). It was also possible to ascertain the topology of Arabidopsis RTNLB13, an ER resident membrane protein of the reticulon family of proteins involved in generating ER curvature (Sparkes et al., 2010). Endomembrane protein topology can also be assessed by exploiting the sensitivity of fluorescent proteins to the environmental pH. For example, YFP fluorescence is sensitive to low pH, a characteristic that can be informative in establishing whether the portion of a target protein to which the YFP is fused is exposed to a low-pH environment, such as the lumen of the vacuole or the apoplast, compared with a reporter fusion for the same target protein, which will be fluorescent only on the cytosolic side of the membrane. Although this approach has been very informative in defining the orientation of important proteins in plant endomembranes, such as the auxin influx carrier AUX1 (Swarup et al., 2004), this kind of approach may not work for those proteins that are sensitive to the attachment of a fluorescent protein tag on either terminus of the protein. Furthermore, the assay requires a steep pH gradient across the membrane. Alternatively, it is possible to use ratiometric redox-sensitive GFP (roGFP) (Hanson et al., 2004; Meyer et al., 2007; Gutscher et al., 2008) and ratiometric imaging (Brach et al., 2009) (Figure 1). The method is based on the difference in the redox potential of glutathione (EGSH) across the ER membrane, which can be sensed by roGFP.

A ratiometric fluorescence measurement can be performed on membrane proteins fused to roGFP to gather direct information on the orientation of the protein with respect to the ER membrane (Brach et al., 2009). With this approach, the orientation of two novel proteins involved in ER export, KMS1 and KMS2, has been established (Wang et al., 2010). The topology of several Arabidopsis reticulons was investigated using roGFP fusions (Sparkes et al., 2010). It has also been possible to show that, unlike non-plant homolog predictions, the H/KDEL receptor has six transmembrane domains (Brach et al., 2009).

The study of plant endomembranes with GFP-based optical tweezers

Optical tweezers have been used to assess processes at the whole cell level down to the nature of molecular interactions, with more recent applications encompassing the physical nature of the cytoplasm and interorganellar connections (Ashkin, 1997; Hawes et al., 2010). For example, optical tweezers have been used to ascertain the requirement for nuclear positioning in tip growth of root hairs (Ketelaar et al., 2002), myosin processivity in tobacco (Tominaga et al., 2003), the potential role of the ER in sensing gravitropism in Arabidopsis (Leitz et al., 2009), requirements for chloroplast division in the red alga Cyanidioschyzon merolae (Yoshida et al., 2006), and in assessing the physical properties of plant cytoplasm (Ashkin and Dziedzic, 1989; Grabski et al., 1994, 1998; Grabski and Schindler, 1995, 1996; van der Honing et al., 2010; Ketelaar et al., 2010).

The principle of optical tweezers employs the radiation pressure generated by a focused beam, usually in the infrared spectrum, being able to ‘trap’ an object with a higher refractive index than the surrounding medium. Upon trapping, the user is able to micromanipulate the object in the X, Y and Z axes, with subsequent movements generating forces in the piconewton range (Neuman and Block, 2004). A single optical trap can capture an object within the micron range; however, multiple traps can be employed to trap larger objects, such as nuclei.

Spatial constriction in highly vacuolated cells results in close associations between organelles. Through the use of optical tweezers, it has been possible to show that associations between Golgi and the ER (Sparkes et al., 2009b) and chloroplast and ER (Andersson et al., 2007) may, in fact, represent a physical connectivity between the two structures. It is unclear at this time whether these represent membranous connections or protein tethers.

The ER and Golgi are functionally linked through the secretory pathway, with post-translationally synthesized protein being subsequently modified in the Golgi. Transport occurs via the anterograde [coat protomer complex II (COPII) mediated] and retrograde (COPI mediated) pathways, allowing passage to the Golgi and retrieval back to the ER, respectively, of resident enzymes (Cai et al., 2007; Robinson et al., 2007; Miller and Barlowe, 2010). Golgi bodies are closely associated with and appear to run over the surface of ER tubules, and the movement of both organelles is an actin–myosin-dependent process (Quader et al., 1987; Liebe and Menzel, 1995; Boevink et al., 1998; Nebenführ et al., 1999; Sparkes et al., 2009a; Ueda et al., 2010). Long-standing questions regarding the ER–Golgi interface include the composition of the connection (membranous tubular connections versus protein tethers), whether it is transient or stable, and whether the movement of the two organelles is interdependent or independent.

To investigate these questions, Sparkes et al. (2009b) employed optical tweezers. They were able to show that, unlike the ER, Golgi bodies could be trapped in Arabidopsis leaf epidermal cells, and the subsequent micromanipulation of the trapped Golgi resulted in remodeling of the ER (Figure 2). These studies therefore highlight that there is tight physical association between the two organelles, but do not infer that the movement of the Golgi drives ER remodeling in vivo. Subsequent Golgi deconstruction studies through the use of brefeldin A (BFA) have shown that in the absence of Golgi bodies the ER is still motile and becomes more cisternal (Sparkes et al., 2009a). While this indicates that Golgi bodies may not drive the majority of ER remodeling in vivo, it does not preclude the possibility that a small proportion could be driven in this way. Future development of tools which deconstruct Golgi bodies without a concomitant cisternalization of the ER, or which disconnect Golgi from the ER, are required before fine mapping of ER dynamics and the potential role of Golgi bodies can be fully elucidated.

Figure 2.

 Optical tweezers highlight the physical connectivity between Golgi bodies and the endoplasmic reticulum (ER) in Arabidopsis leaf epidermal cells.
Arabidopsis leaf samples were treated with latrunculin B to perturb organelle movement. Upon trapping and moving a Golgi body (arrow) the ER was seen to trail behind the Golgi body. The ER could also be ‘wrapped’ around static islands of ER (circles), termed putative anchor points, to form a stable geometric network. Images were taken from a time lapse movie with the time shown in seconds. Scale bar = 2 μm.
Reprinted from Sparkes et al., 2009b, Copyright © 2009 John Wiley & Sons.

Upon depolymerization of the actin cytoskeleton, the ER network produces more blunt-ended tubules. Using optical tweezers, it was shown that Golgi bodies attached to an ER tubule could then dock onto and fuse with a blunt-end ER tubule (Sparkes et al., 2009b). This interaction appeared to require a minimum time frame for association with tethers/tubule fusion to occur. Occasionally, it was possible to rip a Golgi free of the associated ER and to then dock and reassociate it with the ER. Potential candidates for Golgi–ER tethers are myosins and Golgi matrix proteins (Hawes et al., 2008).

The Arabidopsis myosin gene family contains 17 myosins, which fall into classes VIII and XI (Reddy and Day, 2001). Class XI is similar to class V and is required for organelle movement. Through T-DNA insertion, RNA interference (RNAi) and overexpression of fluorescent myosin truncations, it has been shown that there are several class XI myosins (XI-C, XI-E, XI-I, XI-K, XI-1, XI-2) required for global organelle movement in Arabidopsis and tobacco (Avisar et al., 2008, 2009; Peremyslov et al., 2008; Sparkes et al., 2008; Sparkes, 2010; Ueda et al., 2010). Therefore, it is unclear at this time whether a specific myosin is required for independent or dependent movement of the ER and Golgi, or tethering between the two compartments. Plant Golgin homologs have been shown to locate to Golgi bodies, and time-resolved studies indicate that recruitment occurs early during re-formation, indicating a possible role in Golgi formation (Latijnhouwers et al., 2005a,b, 2007; Renna et al., 2005; Kang and Staehelin, 2008; Schoberer et al., 2010). It remains to be understood what specific role Golgins play, but roles in the formation of stack polarity, tethering between stacks or to the ER, or forming a scaffold for Golgi formation have been suggested (Schoberer et al., 2010).

Other than showing a tight physical association between the ER and the Golgi, the optical tweezer studies highlighted another fascinating aspect related to ER network formation/geometry and putative anchor points. Upon depolymerization of the actin cytoskeleton, occasional small, isolated cisternal islands of ER were generated. Micromanipulation of the ER through the movement of trapped Golgi bodies showed that ER tubules could essentially be ‘wrapped’ around these small cisternal islands and subsequently anchored to form a stable geometric structure (Figure 2) (Sparkes et al., 2009b). An independent study by the same group using an approach to map ER dynamics again identified small static regions in the ER network (Sparkes et al., 2009a). Based on these two studies, it was speculated that these static nodes are anchor points holding the dynamic ER network to the plasma membrane to either provide a certain level of stability in maintaining the dynamic ER network throughout the entire cortex, and/or in a signaling capacity to mediate fast transfer of external stimuli sensed by the plasma membrane to the ER. Candidates for anchor point complex formation include class VIII myosins, some of which are prevalent on the plasma membrane, with ATM1 being reported as co-locating with the underlying ER (Golomb et al., 2008).

Chloroplasts are interconnected by long tubular emanations called stromules (Hanson and Sattarzadeh, 2008). Early observations highlighting these lumenal connections include monitoring the movement of GFP from one chloroplast to another (Kohler et al., 1997; Knoblauch et al., 1999). Similarly, the use of optical tweezers in spinach leaves highlighted that a predicted force of more than 70 pN was required to disassociate a trapped chloroplast from the surrounding chloroplasts (Bayoudh et al., 2001). Quantitative studies of the intricate network of stromules and their correlation with the ER network in various Arabidopsis and tobacco cell types using fluorescent fusions has indicated a close association (Schattat et al., 2011). It is unclear, though, whether correlations between stromule and ER branching represent cytoskeletal interactions or a physical association between the two compartments. Using optical tweezers, a physical association was shown by Andersson et al. (2007), who were able to micromanipulate ER associated with isolated chloroplasts from lysed protoplasts. Interactions were reported to be in the 400 pN range, but have yet to be shown in intact cells.

Green fluorescent protein-based forward genetics screens of plant endomembranes

Green fluorescent protein and forward genetics in Arabidopsis

We are learning much about how plant endomembranes work through the functional characterization of plant homologs of factors that are known to play a role in the activities of endomembranes in non-plant cells (Robinson et al., 2007). However, the presence of plant-specific organelles and of ubiquitous organelles with plant-specific features makes it very difficult to fully understand how plant endomembranes establish and maintain their morphological and functional identity by relying exclusively on knowledge from other kingdoms. The opportunity to employ GFP as a visual reporter for the endomembrane system coupled with the genomics tools that are available for the model species Arabidopsis thaliana has led to the development of successful forward genetics screens for the identification of mutants of the secretory pathway and the endocytic route. These screens may lead to the identification of plant-specific factors that control the endomembranes; however, they may also overcome limitations associated with reverse genetic approaches for the identification of molecular components of membrane traffic in plant cells, such as functional redundancy and/or lethality of the mutants. They also allow for the identification of weak alleles of important trafficking regulators, which would not be identifiable through an analysis of morphological phenotypes.

These screens are based on fluorescence microscopy analyses of segregating mutants derived from ethyl methanesulfonate (EMS) mutagenized parental lines expressing fluorescent protein-based reporters of secretory organelles and extracellular compartments, such as the apoplast. The screens are based on the identification of plants showing localization of the reporters in compartments and structures that are different from the wild type (Figure 3). By introducing into the mutants fluorescent markers of organelles and trafficking routes by either simple floral crosses or stable or transient transformation, it is possible to expand the findings from the identification of a mutated gene to the discovery of the function of its product in the secretory pathway or endocytic routes. For a detailed description of this methodology, we refer the reader to a recent publication (Stefano et al., 2011a). Because of the size of the screening population in which samples are inspected individually with laser scanning confocal, fluorescence or stereo microscopes, the screens may be labor intense. However, they can be very rewarding, as reported in the few examples below. Furthermore, the employment of next generation sequencing for mapping the mutation is considerably reducing the time and costs involved in fine mapping using classical approaches (Austin et al., 2011; Marti et al., 2010; Schneeberger and Weigel, 2011; Stefano et al., 2011a). It has been possible to identify new genes involved in the morphology and activities of the secretory pathway, but also to attribute unexpected and novel functions to already characterized genes. These types of screens generally require a well-trained eye and a good dose of patience and determination, especially for identifying mutants with subtle yet significant changes in the distribution of organelle markers from large segregating populations. However, recent exciting advances in automated microscopy (Salomon et al., 2010; Drakakaki et al., 2011) could allow, at least in some instances, a faster identification of mutants.

Figure 3.

 Examples of mutants that can be isolated in forward genetics screens of plant endomembranes.
Confocal images of non-mutagenized Arabidopsis cotyledons stably expressing either a soluble endoplasmic reticulum (ER) marker (A, ssGFP-HDEL) or a Golgi marker (C, ST-GFP). Arrows indicate ER and Golgi A and C, respectively. In B and D, arrows indicate aberrant structures where the markers are localized in isolated EMS mutants (B, ssGFP-HDEL; D, ST-GFP). Scale bars = B = A: 10 μm; D = C: 5 μm.

Identification of mutants of secretory organelles

Plant endomembrane-related functions for gene products that would not otherwise have been identified by bioinformatics searches can also be ascribed. Because EMS generally causes point mutations in the genomic DNA, it is possible to isolate missense mutants with partial or complete loss of gene function. In a case for which a complete loss of function may be lethal, the EMS mutants may overcome some of the limitations associated with the study of essential genes.

For example, in two independent screens for either Golgi defects (Boulaflous et al., 2008), based on the analysis of the subcellular distribution of the Golgi marker ST-GFP (Boevink et al., 1998), or ER morphology defects, through the analysis of the distribution of the lumenal marker GFP-h (Nakano et al., 2009), a mutant, dubbed G92/ERMO2, was identified by the presence of globular structures (Faso et al., 2009; Nakano et al., 2009). These structures were visualized through the partial distribution of ST-GFP (Faso et al., 2009) to the underlying ER or through the abnormal distribution of the ER marker (Nakano et al., 2009). The ER appeared partially deformed into a skein-like structure in the perinuclear area and in smaller globular structures scattered within the ER network.

The mutation was mapped to a missense mutation in AtSEC24A (Faso et al., 2009; Nakano et al., 2009), one of the three Arabidopsis isoforms encoding the COPII protein SEC24 that is involved in ER-to-Golgi protein transport (Barlowe et al., 1994; Robinson et al., 2007). This mutation caused an amino acid switch of an arginine to a lysine residue (Faso et al., 2009; Nakano et al., 2009) in a highly conserved site involved in cargo selection for ER export (Miller et al., 2003; Mossessova et al., 2003), and it partially compromised the distribution of AtSEC24A to the ER exit sites (Faso et al., 2009), suggesting that the G92/ERMO2 mutation was linked to a partial loss of function of ASEC24A.

Interestingly, the mutation did not cause a visible plant phenotype; however, T-DNA knockout mutants of AtSEC24A could not be recovered (Faso et al., 2009; Nakano et al., 2009). This suggested that while G92/ERMO2 is linked to a partial loss of function, AtSEC24A mostly likely shares partially overlapping roles with the two other AtSEC24 isoforms; however, the complete loss of AtSEC24A is lethal. This was later confirmed by the evidence that a complete loss of AtSEC24A caused defects in pollen and male sterility (Conger et al., 2011).

The characterization of G92 also revealed that the AtSEC24A mutation caused partial inhibition of secretion of bulk-flow and membrane markers at the globular non-Golgi structures marked by ST-GFP (Faso et al., 2009). These findings led to the proposition that partial loss of function of AtSEC24A could influence the trafficking of specific cargo necessary for the integrity of the ER, but also ER export in a region of the cell (Faso et al., 2009; Nakano et al., 2009). Therefore, the findings led to new insights in plant COPII components, not only from a genetic perspective, but also due to the suggestion that AtSEC24 proteins may have a spatially regulated activity in the cells.

Identification of mutants of secretory routes

The advantage of using GFP-microscopy screens for plant endomembranes is that it is also possible to follow anomalies both in the organelles, where the reporter is targeted, and/or in intermediate structures through which the reporter may traffic. For example, in two independent screens for defects in the distribution of either the vacuolar marker GFP:δ-TIP, a tonoplast marker (Avila et al., 2003), or ST-GFP, a Golgi marker, mutants with partial distribution of the reporters in the ER were identified. They were named modified vacuole phenotype1-1 (mvp1-1; Agee et al., 2010) and Golgi defective 36 (gold36; Marti et al., 2010) in the screen for the tonoplast and Golgi marker, respectively. In both screens, comparative co-localization analyses with endomembrane markers in the mutant backgrounds and the wild type showed that the ER was abnormal with the presence of globular and ring-like structures, in which soluble bulk-flow and vacuolar markers were also distributed. Interestingly, the mutations were mapped to an allelic mutation in the At1g54030 genetic locus coding a small protein (Agee et al., 2010; Marti et al., 2010), which is generally exported to the vacuole lumen (Marti et al., 2010).

These findings highlight that via a screen focused on a specific organelle, it is possible to identify defects of other organelles through which the GFP marker is partially distributed. In addition, they support the possibility that the integrity of organelles may depend on proteins that do not reside in the organelle affected by the mutation. Interestingly, it was also found that an At1g54030 knockout mutant phenocopied the phenotype of the mvp1-1 and gold36 alleles (Agee et al., 2010; Marti et al., 2010). Based on these findings, the interesting hypothesis was proposed that At1g54030 may encode a factor that is necessary for integrity of the ER because of its ability to limit damage to the ER by chaperoning deleterious proteins away from this organelle (Marti et al., 2010).

Green fluorescent protein-based screens allow the study of conserved genes in the context of plant endomembranes

With GFP-based microscopy screens, it is possible to identify proteins whose function can be deduced based on studies in other systems. However, they also allow increased general knowledge of trafficking components in plant cells, which is comparatively less well known than in yeast and mammalian systems, and to place regulators known in other biological systems into a new functional context. This is exemplified by the results from a screen for aberrant distribution of PIN1pro:PIN1-GFP, an endogenous promoter fusion to GFP-tagged PIN1. This is an auxin efflux carrier, which is an integral membrane protein associated with the plasma membrane at the basal side of stele cells at steady state and undergoes constitutive endocytosis (Dhonukshe et al., 2006, 2007).

Such screens led to the identification of protein affected trafficking (pat) mutants (pat2 and pat4) with partial accumulation of PIN-GFP to intracellular structures (Feraru et al., 2010; Zwiewka et al., 2011). The characterization of these mutants led to the identification of members of AP-3 adaptin complex (Dell’Angelica, 2009). While pat2 was allelic to AP-3 β, pat4 was allelic to AP-3 δ (Feraru et al., 2010; Zwiewka et al., 2011). It was demonstrated that the mutants had several cellular defects in vacuole identity and function, including altered degradation and abnormal protein accumulation in misshapen lytic vacuoles, aberrant vacuolation of pre-vacuolar compartments and impaired transition from storage to lytic vacuoles, but no defects in secretory or early endocytic processes. Furthermore, it was shown that AP-3 β could interact with clathrin, demonstrating that the AP-3 complex exists in plant cells and that the AP-3 complex may have acquired plant-specific functions in regulating biogenesis and the function of vacuoles (Feraru et al., 2010). These findings significantly advanced our knowledge of PIN proteins, thus significantly expanding the understanding from earlier screens based on gravitropic responses that led to the identification of PINs in the first place (Muller et al., 1998). Interestingly, similarly to the G92/ERMO2 mutant described above, the pat2 and pat4 mutants did not have strong morphological plant phenotypes, although the mutations clearly affected important organelles and secretory processes, further highlighting that the GFP screens enable the identification of mutants that could be missed if the screens were based on plant phenotype.

Utilization of the GFP limitations for the success of forward genetics screen

The GFP-based microscopy screens can reach a further level of sophistication by taking advantage of intrinsic limitations of fluorochromes. As GFP is generally sensitive to low pH, it is quite difficult to detect in the apoplast. Furthermore, it has been shown that in compartments such as the vacuole or apoplast, GFP can be cleaved (Tamura et al., 2003; Zheng et al., 2004). To account for the degradation in the vacuole, it was suggested that blue light may induce a conformational change in GFP, which could then be easily degraded by vacuolar papain-type cysteine proteinase(s) under the acidic pH, making it difficult to detect vacuolar GFP in the vacuole of light-grown plants (Tamura et al., 2003). These properties are the basis of two independent screens. In one of these, the distribution of the secreted bulk-flow marker, sec-GFP, a fusion of the sporamin signal peptide to GFP, was followed in roots of EMS-mutagenized individuals using a stereo microscope (Zheng et al., 2004). Mutants with defects in secretion would be scored for the presence of GFP signal in the roots, while plants with a wild-type phenotype would not show a detectable GFP signal. This screen led first to the identification of a mutant allelic to the ROOT HAIR DEFECTIVE 3 allele (rhd3-1), which encodes a non-essential ER-anchored GTPase involved in ER structure (Zheng et al., 2004; Chen et al., 2011; Stefano et al., 2011b). The identification and characterization of the mutant in this work led to the suggestion that RHD3 may be involved in protein secretion as well as ER structure (Zheng et al., 2004). However, recent findings based on the expression of RHD3 dominant-negative mutants with altered conserved residues in the putative GTPase domain do not seem to fully support an involvement of RHD3 in protein secretion, because no inhibition of ER export of soluble and membrane markers was detected in tobacco leaf epidermal cells co-expressing putative RHD3 dominant-negative mutants (Chen et al., 2011). It is yet to be explored, however, whether this apparent discrepancy may be linked to the possibility that the RHD3 dominant-negative mutants were expressed in a heterologous background or/and that the rhd3-1 mutation might cause more severe defects in the ER than the mutation in the putative GTPase domain.

The sec-GFP screen led to additional interesting findings upon the identification and characterization of a gnom-like 1 (GNL1) allele (Teh and Moore, 2007). GNL1 is a member of the GBF ARF-GEF (guanine-nucleotide exchange factors for ADP-ribosylation factor GTPase) superfamily involved in vesicle coat formation and vesicle–cytoskeleton interactions (Geldner, 2004; D’Souza-Schorey and Chavrier, 2006). GNL1 is a close relative of GNOM, a GEF involved in recycling from endosomes to the plasma membrane (Geldner et al., 2003). The characterization of the GNL1 mutant showed that the protein is localized at the Golgi and that it is involved in the biosynthetic anterograde route (Teh and Moore, 2007). Furthermore, the study not only showed that GNL1, unlike its close homolog GNOM1, is resistant to inactivation by the fungal metabolite BFA, but also that the protein has a role in selective endocytosis and in Golgi morphology, since the Golgi is roughly 50% larger in the gnl1 mutant than the wild type (Teh and Moore, 2007). These findings were subsequently extended by the characterization of an additional GNL1 allele, named ERMO1, identified through a GFP-based microscopy screen of Arabidopsis EMS mutants expressing the ER lumenal marker GFP-h (Nakano et al., 2009). In this work, it was demonstrated that the ER structure is also compromised in a gnl1 loss-of-function mutant background, supporting the possibility that GNL1 may have acquired distinct attributes and may control the activation of different GTPases that may act on various secretory organelles.

Similarly to the screen for aberrant distribution of secGFP and the possibility of consequent visualization of the marker in a microscopy screen, the sensitivity of GFP to lytic environments has been exploited in a GFP-based screen of Arabidopsis EMS mutants expressing the vacuole-targeted GFP-2SC (Tamura et al., 2007). Transgenics expressing this marker do not usually show fluorescent vacuoles with growth under light. However, mutants with defective biosynthetic traffic are readily identifiable due to GFP fluorescence in compartments that do not disrupt the fluorochrome. This type of screen has led to the identification of KATAMARI 2 (kam2), an allele to the Arabidopsis gravitropism defective2 (grv2) mutant, which is deficient in shoot gravitropism and phototropism (Silady et al., 2004), and to the Arabidopsis homolog of a DnaJ domain-containing RECEPTOR-MEDIATED ENDOCYTOSIS-8 (RME-8), a mammalian homolog, which is believed to play a role in endocytosis (Fujibayashi et al., 2008). Characterization of kam2 led to the findings that KAM2 has important roles in seeds being involved in the transport of proteins into the protein storage vacuole, as well as in the determination of the growth axis of the embryo (Tamura et al., 2007), thus ascribing plant-specific functional aspects to ubiquitous gene products.

Unexpected players in organelle integrity

In addition to genes which may be implicated in the morphology of secretory organelles by sequence similarities with other genes, GFP screens have allowed the identification of genes that would have not been implicated in this by simple homology searches. For example, an unexpected player in organelle positioning has been identified through the characterization of the KATAMARI1 (kam1) mutant, isolated through a GFP screen for mutants with partial distribution of the vacuole-targeted GFP-2SC in non-vacuolar compartments (Tamura et al., 2005). KAM1 showed the presence of a large aggregate in each cell containing ER membranes. KAM1 was found to be allelic to the MUR3 gene, which encodes a Golgi-localized protein containing a exostosin-like domain responsible for glycosyltransferase activity (Madson et al., 2003). Interestingly, analyses of other mur3 alleles defective in glycosyltransferase activity did not show defects in the organization of the endomembranes, suggesting that the protein has multifunctional properties (Tamura et al., 2005). In support of this, it was also shown that KAM2 interacts with actin, suggesting the possibility that KAM2 is involved in control of the organization of the ER through an interaction with the cytoskeleton (Tamura et al., 2005), a function that would not have been ascribed based simply on the glycosyltransferase activity associated with MUR3.

A GFP screen for components involved in endocytosis

A GFP-based imaging screen has also been developed to identify components involved in endocytosis, a process for which very important pieces of information are lacking in plants. To do so, a screen was developed for the identification of mutants defective in internalization of proteins that recycle from the plasma membrane, such as the PINs. In particular, PIN1 is known to reside at the plasma membrane; however, in the presence of BFA, PIN1 is partially redistributed to compartments, known as BFA compartments which contain mainly endosomes and are formed as a consequence of the inhibition of BFA-sensitive ARF-GEFs involved in recycling to the plasma membrane (Geldner et al., 2001; Dhonukshe et al., 2007). PIN1 cycling to the plasma membrane is sensitive to BFA, but endocytosis is generally insensitive to BFA. Therefore, the screen for reduced accumulation of PIN1 into the BFA compartments would report on defects in endocytosis (Tanaka et al., 2009). This screen has been particularly fruitful for the identification of players in endocytic processes with the identification of BEN1/AtMIN7/BIG5, a member of the BIG subclass of ARF-GEFs, as a GEF working at the level of TGN/early endosomes (Tanaka et al., 2009).

‘Side-effects’ with a happy ending for GFP-based screens

Finally, but no less importantly, GFP-based screens can lead to the serendipitous isolation of mutants that do not affect the endomembranes. Because the screens are based on microscopy analyses of the distribution of markers that can define the cell contour, such as tonoplast markers, it is possible to identify mutants with defective organization of their cell pattern and growth. This has been the case, for example, of the identification of a mutant named cell shape phenotype-1 (csp-1), an allele of AtTPS6, which clearly showed defects in cell patterning. The cell defects in the AtTPS6 allele were visualized through the distribution of a tonoplast marker by fluorescence microscopy (Avila et al., 2003; Chary et al., 2008). Through the characterization of csp-1, it was possible to establish that defects in AtTPS6, which is thought to be involved in the synthesis of trehalose, lead to phenotypes in cell morphogenesis and but also in plant growth and development (Chary et al., 2008), thus adding important insights linking sugar metabolism and plant development.


We are grateful to Ms Eileen Morey for editing the manuscript and to Dr Luciana Renna and Ms Lucia Marti for providing the images in Figure 3. We acknowledge support by the Chemical Sciences, Geosciences and Biosciences Division, Office of Basic Energy Sciences, Office of Science, US Department of Energy (award number DE-FG02-91ER20021), National Science Foundation (MCB 0948584) (FB), Oxford Brookes University, BBSRC and the Leverhulme Trust for funding some of the work reported in this review.