Multidimensional fluorescence microscopy of multiple organelles in Arabidopsis seedlings

Background The isolation of green fluorescent protein (GFP) and the development of spectral variants over the past decade have begun to reveal the dynamic nature of protein trafficking and organelle motility. In planta analyses of this dynamic process have typically been limited to only two organelles or proteins at a time in only a few cell types. Results We generated a transgenic Arabidopsis plant that contains four spectrally different fluorescent proteins. Nuclei, plastids, mitochondria and plasma membranes were genetically tagged with cyan, red, yellow and green fluorescent proteins, respectively. In addition, methods to track nuclei, mitochondria and chloroplasts and quantify the interaction between these organelles at a submicron resolution were developed. These analyzes revealed that N-ethylmaleimide disrupts nuclear-mitochondrial but not nuclear-plastids interactions in root epidermal cells of live Arabidopsis seedlings. Conclusion We developed a tool and associated methods for analyzing the complex dynamic of organelle-organelle interactions in real time in planta. Homozygous transgenic Arabidopsis (Kaleidocell) is available through Arabidopsis Biological Resource Center.


Background
Advances in fluorescence microscopy [1] and the development of multiple spectral variants of the green fluorescent protein (GFP) [2] have revolutionized plant cell biology. These technological developments have produced a wealth of published information related to protein localization, and are currently being exploited to analyze protein-protein interactions [3]. Recently, this fluorescence technology has begun to be applied to an analysis of the dynamic motility of several organelles, including mitochondria [4][5][6] nuclei [7] and Golgi [8]. Much of this information has been deposited in the plant organelle database (PODB) [9]. However, the majority of these plant studies have been performed using transient assays or have involved marking only a few organelles with one or two fluorescent proteins, making it difficult to perform detailed analyses of the interaction or association between multiple organelles. In both plants and animals, some organelles are localized in close proximity to one another such that the distance that separates them is shorter than the resolution limit of a conventional fluorescence microscope (< 300 nm), but longer than the resolution limit of a conventional electron microscope (> 20 nm). In the context of the studies presented here, we describe these close localizations of organelles as organelle-organelle interactions (OOIs).
A classical example of OOIs in plants are those involving chloroplasts, peroxisomes and mitochondria in green tobacco leaves, observed by electron microscopy [10]. Because enzymes required for photorespiration are stored separately in these organelles, secondary metabolites, such as glycolate and glycine, must be transported among these organelles. Hence, in this case, the OOIs may facilitate photorespiration by maintaining short transportation distances. Using confocal microscopy, physical interactions between the nucleus and plastids were recently described in tobacco hypocotyl epidermal cells [11], and signal transduction between the nucleus and plastids has also been reported [12]. Kwok and Hanson have hypothesized that nucleus-plastid signaling may be facilitated by OOIs that reduce the distance between the nucleus and plastids [11]. In animal cells, OOIs between the nucleus and mitochondria have also been described [13]. In a study of ATP transport, physical interactions between nuclei and mitochondria were observed in cultured cells using both confocal and electron microscopy. Although nuclei are not capable of producing ATP, a large amount of it is required for nucleus-specific functions, such as transcription and mRNA transport. Thus, nucleus-mitochondria OOIs may facilitate ATP transport from the mitochondria to the nucleus [13].
Although biologically important roles for OOIs have been hypothesized as described above, the nature of organelleorganelle physical interactions is poorly understood. Indeed, molecules that mediate the physical interactions between nuclei and plastids or mitochondria have yet to be identified. Organelles may interact with each other through an active molecular recognition mechanism or perhaps, if cellular space is limited, through a passive process. In this study, we sought to determine which phenomenon -active or passive recognition -is correct by addressing the following questions: 1. How many plastids and mitochondria are in each cell?
2. How much space is available in a cell for the nucleus, plastids and mitochondria? 3. Where do these organelles locate in the three-dimensional (3-D) space of the cell? 4. How long do the OOIs last?

Are specific molecular interactions involved in mediating OOIs?
To begin to answer these questions, we have developed a transgenic Arabidopsis plant as a tool for analyzing the dynamics of multiple organelles and multiple OOIs in real time in planta.

Establishing homozygous Kaleidocell lines
Transgenic Arabidopsis plants, designated Kaleidocell (in Greek, kalos beautiful + eidos form + cell), were engineered to contain multiple fluorescent proteins that label organelles in vivo. Specifically, nuclei, plastids, mitochondria and plasma membranes were genetically tagged with cyan fluorescent protein (CFP), red fluorescent protein (RFP), yellow fluorescent protein (YFP) and green fluorescent protein (GFP), respectively ( Figure 1A). The fluorescent proteins and tagged peptides used in this study have been shown previously to be sufficient to direct fusion proteins to each cellular component in plants. Nuclei were tagged with ECFP (Clontech, CA) fused to the yeast nuclear localized protein, Gal4 [14], and the nuclear localization signal from SV40 large T-antigen [15]. Plastids were tagged with DsRed2 (Clontech) fused to the Arabidopsis RecA signal peptide [16]. Mitochondria were tagged with EYFP (Clontech) fused to the mitochondrial localization signal sequence in yeast cytochrome oxidase IV [17]. Plasma membranes were tagged with sGFP [18] fused to the petunia calmodulin CaM53 [19,20].
To confirm transgene integration and to predict transgene copy numbers, we performed Southern blots on BamHIdigested genomic DNA from homozygous Kaleidocell using a probe encoding the promoter region of each of these transgenes ( Figure 1B). We detected four bands, indicating that a single copy of each expression cassette (CFP, RFP, YFP, and GFP) was integrated in the Kaleidocell genome. One of the bands appeared at approximately 1.4 kbp, which is 2.0 kbp shorter than expected-shortest size (3.4 kbp). The reduced size of this band suggests that one of the T-DNAs did not integrate completely into the genome. However, the fact that all four spectral variants were visible in Kaleidocell specimens indicates that this truncation of T-DNA had no apparent effect on the functionality of the four transgenes.
To confirm proper accumulation of each fluorescent protein, protoplasts were generated from one-week-old homozygous Kaleidocell seedlings ( Figure 1C). Within a single protoplast, we detected cyan, red, yellow, and green fluorescence in the nucleus, plastids, mitochondria, and plasma membrane, respectively. The fluorescence spec-trums of CFP and GFP overlap with one another, as do those of GFP and YFP. Accordingly, it can be challenging to distinguish CFP, GFP and YFP signals in a single cell [21]. Although these signals can be separated through the use of spectral imaging and linear un-mixing of image data [22], the microscopes and software used in this study did not allow us to do so. Furthermore, the petunia cal-modulin CaM53 that tags plasma membranes tended to accumulate in nuclei by blocking isoprenoid biosynthesis or depleting carbons [19]. To distinguish organelles from one another in this study, we used a combination of fluorescent signals and organelle size. The nucleus was a clearly recognizable entity with a diameter larger than 5 μm. Meanwhile, plastids were entities with diameters just Transgene constructs and their expression in the Kaleidocell line larger than 2.5 μm, and mitochondria were the smallest recognizable objects with a diameter greater than 0.7 μm.
The fluorescent proteins in Kaleidocell seedlings (3 days to 2 weeks post-germination) were easily detected in the hypocotyls and roots of all individuals analyzed ( Figure  2A and 2B). However, three weeks post-germination, the fluorescent protein signals were not as readily detected. Although the reason for this time-dependent reduction in signal intensity is unknown and requires further study, it is likely that the transgene loci may be affected by local chromatin modifications, which alter transgene expression levels at distinct stages of plant development.
Because Kaleidocell is homozygous for all four fluorescent proteins, plants can be propagated without using a selection reagent such as kanamycin. Additionally, we have found that the fluorescence of all four fluorescent proteins remains stable in subsequent generations (data not shown).

Mounting living seedlings on a microscope stage
Living plants can be adhered to a coverglass for a number of hours using a gelatin-or agar-based gel mounting medium [23]. Although this method generally works well, it is not suitable for high-resolution 3-D microscopy because of optical scattering and aberrations caused by the gel (Kato unpublished). To optimize the quality of the signal captured from the Kaleidocell plant, we developed a simple method that allowed us to make extended obser-vations (hours) of whole seedlings with minimum optical scattering or other aberrations. Our method uses watersoaked Kimwipes to provide 1) a mounting medium (water) that fills the empty space between the sample and the coverglass and reduces dehydration of the sample being analyzed, and 2) a force that gently pushes the seedling towards the coverglass ( Figure 3A-C). During our experiments, Arabidopsis seedlings mounted in this manner grew for over 12 hours on a microscope stage (data not shown).
In addition, a convenient method also was developed to enclose Kaleidocell in a microscope slide ( Figure 3D). After germinating seeds for 3 days on water-soaked blotting paper in a plastic Petri dish, a microscope slide with a one-well adhesive spacer (20 mm in diameter, 0.12 mm deep) was used to mount Kaleidocell with water. A glass cover slip was used to seal a seedling in this chamber.
Although it may be difficult to continuously observe single cells for extended periods due to the gradual movement of Kaleidocell in water, it is possible to obtain sequential high-resolution images over a short period of time (e.g., 10 minutes). This method may be useful to quickly test or demonstrate the feasibility of Kaleidocell.

Detection of nuclei, plastids and mitochondria in epidermal and cortical cells of the Kaleidocell root
To visualize the multiple fluorescent proteins marking the different organelles in Kaleidocell, we obtained a stacked confocal scanning microscopic image of a Kaleidocell Expression of the transgenes in seedlings in the Kaleidocell line seedling root (Figure 4). The resolution of the objective lens used in this scan was calculated at 0.25 μm in the xand y-axes and 0.65 μm in the z-axis. Because of the limitations of our laser lines, we acquired GFP (plasma membranes) and YFP (mitochondria) signals in the same channel. However, as noted above, the size and pattern of each of these organelles made it easy to distinguish the plasma membrane from the mitochondria.
A total of 25 cells, 17 epidermal and eight cortical, were clearly identified within a 230.2 × 230.3 × 30.9 μm scan area ( Figure 4). Of the 25 identified cells, 11 nuclei were inside the scan area during acquisition. The number of plastids and mitochondria were determined using imaging software. Point-like structures (spots) were automati-cally identified, and 255 RFP spots (diameter > 2.5 μm) and 10,158 GFP/YFP spots (diameter > 0.7 μm) were detected. Hence, we estimated that a single cell in the root of the Kaleidocell seedling contained approximately 10 plastids and 400 mitochondria. Mitochondria do not exist in cells as discrete units and could be smaller than our microscopic resolution [4,5], so the actual number of mitochondria may be higher or lower than our reported numbers. However, the number suggested by our analysis is similar to that predicted by the amount of mitochondrial DNA in animal cells [24], providing a measure of confidence that the number of mitochondria we were able to visualize in a single cell would be close to the actual number.
Mounting a seedling on an inverted fluorescence microscope

Three-dimensional positioning of nuclei, plastids and mitochondria
The stacked confocal scanning microscopic image was converted to a 3-D model and visually inspected using imaging software to determine the location of organelles in space ( Figure 5 and see Additional file 1). We found that almost all organelles were located in the cortical region of the cells. By manually identifying cellular locations of nuclei, plastids and mitochondria, we estimated that 60 ± 12.5% (n = 10) of the cell volume did not contain these organelles. This suggests that the cytosolic area in which nuclei, plastids and mitochondria can move comprises approximately 40% of the total cellular volume in epidermal and cortical cells of Arabidopsis roots. However, it has been demonstrated previously that expanded vacuoles in root cells [25] push nuclei, plastids and mitochondria to the cortex. Thus, the total cellular volume available to these organelles may vary as a function of vacuole size and/or complexity in different cells or cell types.

Dynamics of nuclear interactions with plastids and mitochondria
To simultaneously investigate the dynamics of OOIs between nuclei and plastids, and nuclei and mitochondria, we tracked the movements of each organelle in a single root epidermis for 3 h, at 1 min intervals ( Figure 6, Table 1, and see Additional file 2).
During this time, we observed nuclear movements similar to those reported previously [7]. For example, one nucleus moved constantly at an average speed of 2.4 μm/min for 70 min covering a distance of 122 μm; another nucleus within the same tissue did not move as far (4-μm travel distance over 70 min, Figure 6 and see Additional file 3). A previous study indicated that nuclear movements in Arabidopsis root hairs correlated with cell growth [26]. Hence, the growth status of individual cells in the root epidermis may be reflected in nuclear movements.
Plastid movements have also been previously reported to vary [6]. We found the plastids that interacted with nuclei moved with the nuclei over the 3-h observation interval ( Figure 6 and see Additional file 4), indicating that the nucleus constantly interacts with the same population of plastids, independent of their movement within the cell.
The mitochondria were smaller and moved faster than the other organelles, making them difficult to track individually. The software we used was unable to track the movement of the mitochondria for an extended period (Table  1). However, we did observe that the mitochondria that interacted with and surrounded the nuclei moved in the same direction as the nuclei (Figure 6 and see Additional file 5). Although this does not conclusively demonstrate that the nucleus was interacting with the same population of mitochondria, it does suggest that the nucleus is constantly interacting with mitochondria.

Pharmacological analysis of OOIs between nuclei and plastids or mitochondria
Unlike most other eukaryotes, actin microfilaments control organelle movements in higher plants (reviewed in [27,28]). Actin microfilaments, together with myosin motor proteins, also control cytoplasmic streaming [27]. It is also known that microtubules, which control organelle movements in other eukaryotes, contribute to the positioning of organelles in higher plants [28].
To explore the nature of the nucleus-plastid and nucleusmitochondria OOIs, we performed a pharmacological screen to search for factors that might alter these OOIs. We exposed Kaleidocell seedlings to five different pharmacological compounds that have been used previously to study the positions and movements of nuclei, mitochondria or plastids in higher plants [7,[29][30][31]. The first compound, Latrunculin B, decreases the amount of polymerized actin and immobilizes nuclei [7] and mitochondria [29]; it also disrupts plastid positioning [30]. The second compound, oryzalin, influences tubulin polymerization and disrupts the positioning of mitochondria [29]. The third compound, 2,4-dinitrophenol (DNP), accelerates depletion of ATP from the cells and immobilizes mitochondria [29]. The fourth compound, 2-monoxime (BDM), is a mammalian myosin ATPase inhibitor that immobilizes mitochondria and also causes immobilized mitochondria to aggregate [29]. The fifth compound, N-ethylmaleimide (NEM), causes alkylation of proteins and immobilizes nuclei [7], mitochondria [29] and plastids [31], and also causes immobilized mitochondria to aggregate [29].
Nucleus-mitochondria interactions and nucleus-plastid interactions were manually identified under a fluorescence microscope using size, morphology and spectral fluorescence to identify each type of organelle (Figure 7). In the absence of pharmacological compound exposure, 71% of nuclei were observed to interact with mitochondria (n = 128, 6 independent experiments). This percentage was lower than that obtained using confocal imaging (~100% of the observed nuclei), suggesting that the lower resolution of the objective lens used on the epifluorescence microscope may limit the ability to identify all nucleus-mitochondria OOIs. Alternatively, the inability to acquire 3-D images may have been an issue, resulting in mitochondria located beneath nuclei being missed during OOI detections. However, this approach was capable of detecting a majority of nucleus-mitochondria OOIs and the quantitative differences did not interfere with our ability to analyze the effects of pharmacological compounds on these OOIs. To the extent that molecules involved in nucleus-plastid and/or nucleus-mitochondria OOIs are affected by a given pharmacological compound, the prediction is that fewer OOIs would be observed between these organelles in cells treated with the corresponding compound. No significant change in the percentage of nucleus-mitochondria OOIs was observed in seedlings treated with latrunculin or oryzalin compared to mock-treated seedlings ( Figure 7A). These results suggest that cytoskeletal proteins, such as actin and microtubules, do not participate in this type of OOI. In contrast, NEM significantly reduced the percentage of nucleus-mitochondria OOIs (0.3-fold, P ≤ 0.01), suggesting that molecules sensitive to NEM-mediated alkylation may be involved. Moreover, dual exposure to NEM and latrunculin or NEM and oryzalin did not induce a further reduction in the interaction rate beyond that of NEM alone, indicating the absence of a synergistic effect. It is impossible to predict which molecules might control the nucleus-mitochondria OOIs on the basis of this experiment because NEM immobilizes many other organelles and impacts fluid dynamics (cytoplasmic streaming) in plant cells [31]. A more extensive analysis, perhaps involving random mutagenesis of the Kaleidocell line, will be required to identify the molecules involved in nucleusmitochondria OOIs. Analyses of nucleus-plastid OOIs in these same experiments did not reveal any changes in the presence of any of these pharmacological compounds ( Figure 7B). Although these results do not allow us to form any conclusions about the molecular aspects of nucleus-plastid OOIs, they do suggest that molecules or mechanisms other than those involved in nucleus-mitochondria OOIs are responsible for regulating nucleusplastid OOIs. The difference between nucleus-mitochondria and nucleus-plastid interactions also suggest that these interactions are active interactions and not passive interactions.

Conclusion
We demonstrate that Kaleidocell is a useful tool to begin the exploration of OOIs. We have identified nucleus-chloroplast and nucleus-mitochondria interactions in root epidermal cells, and have demonstrated that NEM disrupts nucleus-mitochondria OOIs but not nucleus-plastid OOIs. These results suggest that different mechanisms are involved in regulating nucleus-mitochondria and nucleus-plastid OOIs. Kaleidocell should also be useful in experiments designed to determine if cell type, developmental stage and/or exogenous stimuli are capable of regulating or altering these interactions. The homozygous Kaleidocell line is available through Arabidopsis Biological Resource Center with the stock number CS16303.

Fluorescent proteins to tag plastids
The EYFP region of pEYFP-C1 vector (Clontech) was removed with NheI/SacI and subcloned into pBluescript SK+ (Stratagene) at XbaI/SacI to form pBSK+-EYFP. This vector was subsequently digested with BamHI/SacI and the EYFP containing region was subcloned into pBI221 (Clontech) digested with the same restriction enzymes, forming the pBI221-EYFP vector. A 300 bp fragment, corresponding to the signal peptide of the Arabidopsis RecA protein that localized in chloroplasts (NM_106556) was Effects of pharmacological compounds on nucleus-mitochondria and nucleus-plastid interactions PCR amplified from an Arabidopsis cDNA library using primers LM116 (5' CgggATCCATggATTCACAgCTAgTCTT-gTC 3') and LM117 (5' gAAgATCTTCCATAgCT-gCCTCTAAAgCCTT 3'). This fragment was cloned into the pGEM-T vector (Promega) and sequenced. The RecA signal peptide was subsequently removed from this vector using the restriction sites that were incorporated in the primers (BamHI/BglII) and subcloned into the BamHI site of pBI221-EYFP forming the pBI221-RecA::EYFP vector. The EYFP region of this vector was replaced with the red fluorescent protein DsRed2 (Clontech) at the AgeI/ SacI sites, forming the pBI221-RecA::DsRed2 vector. The region containing RecA::DsRed2 was then transferred to pEL103 [21] at the BamHI/SacI sites, forming the pEL103-RecA::DsRed2 vector.

Plant screening and crossing
Zygosities of each transgenic line were determined by their sensitivity to kanamycin selection and by detection of the fluorescent proteins in their seedlings.

Plant growth conditions to establish the Kaleidocell line
The seeds were germinated on 0.75% agar plates containing a half-strength (1/2) MS salt (Sigma, CA) containing 100 mg/L of kanamycin after sterilizing in 50% diluted Crolox (Crolox, CA). The plates were kept in an environmental chamber in which a temperature was set at 21°C and a day cycle was set at 16 hours light and 8 hours dark.
The plants selected were transplanted in soil and grown in the same chamber.

Southern blotting
The Southern blotting was conducted based on the previously published method [15]. Hydroponically cultured homozygous Kaleidocell was used to extract the genomic DNA.

Protoplast preparation
The protoplasts were prepared based on the previously published protocol [33]. About 50 seedlings of the oneweek-old Kaleidocell line were used as starting material.

Seedling preparation for the extended-hour observation
An one-weeks-old seedling was transferred to a chambered #1.5 cover glass (Lab-Tek II, Catalogue number 155360, Nalge Nunc International, IL) and 4 to 5 sheets of 11 × 21 cm water-soaked Kimwipes ® (Kimberky-Clark, GA) were folded and piled on the top of the seedling in the chamber so that a height of the piled Kimwipes were slightly (i.e., ~5 mm) higher than that of the chamber. A slide (75 mm × 22 mm, Catalogue number 2948, Corning, NY) was used as a lid for the chamber. The slide-covered chamber was then fixed on a microscope stage using labeling tape (1.9 mm width, Catalogue number 15-959, Fisher Scientific, PA) so that the slide could push the water-soaked Kimwipes towards the seedling being analyzed.

Sample preparation for the pharmacological analysis
One-week-old seedlings were transferred to a 60 × 15 mm Petri dish filled with 10 ml of distilled water. Dimethylsulfoxide (DMSO) or water solution containing pharmacological compounds was added into the Petri dish and gently shaken for 1 h at about 15 rpm on a rotory shaker. The seedlings were first transferred to a new Petri dish filled with distilled water to rinse the compounds and then transferred to a slide (75 mm × 22 mm) with a 0.12 mm deep spacer (Catalogue number S24736, Invitrogen, CA). Distilled water was added in the space formed by the spacer and this was sealed with a #1.5 cover glass (Catalogue number 2640, Corning, NY).