Generation of these cell lines involved several steps as described below.

Cloning of erd14/10 with an inducible promoter region (regulated by estrogen) was performed using the same pENTR constructs obtained above. The destination vector pDEST contains a region for promoter binding upon induction and together with the gene of interest it was integrated into BY-2 facilitated by A. tumefaciens infection as described above. The expression profile for either ERD14 or ERD10 in BY-2 and also in Arabidopsis cells growing in suspension was monitored for a period of 96 h and detected by Western Blotting. The labeling medium for incorporating 15N isotopes into BY-2 was prepared by replacing the corresponding inorganic salts (ammonium sulfate and ammonium nitrate) in normal Murashige-Skoog media (Murashige and Skoog, 1962) with the salts incorporating the ¹⁵N isotope.

For localization in planta, ERD14/10 was transiently transformed on Nicotiana benthamiana leaves. A. tumefaciens LBA4404 transformed with the erd14-rfp or erd10-rfp fusion gene was grown in 5 ml YEB medium, supplemented with the corresponding antibiotics at 28°C for 48 h shaking at 180 rpm. OD₆₀₀ was measured and cultures were diluted to a final volume of 4 mL to an OD₆₀₀ = 1.5 and spun down for 5 min at 4,000 g. The pellet was resuspended in 4 mL of infiltration buffer using conical-bottom 15 mL tubes (10 mM MgCl₂, 10 mM MES buffer pH = 5.6 and 0.1 mM acetosyringone). Each tube was incubated for 2 h at room temperature in a tube rotator. Infection was performed by injecting 0.5 mL of the previously described mix on the underside of several N. benthamiana leaves using a syringe without needle (Li, 2011). Upon infection and incubation for 72 h, plants were investigated by confocal microscopy to verify RFP-fused protein production by cutting a section of a leave and placing it directly on a cover slip (section of 1 cm² approx.).

E. coli BL21(DE3) cells were transformed with ERD14-C186 pDNA according to standard procedures and a positive colony was selected from an LB-agar plate supplemented with carbenicillin to prepare a pre-culture of 50 mL liquid NZYM medium (10.0 g/L NZ amine, 5.0 g/L NaCl, 5.0 g/L Bacto-yeast extract, 2.0 g/L of MgSO4.7 H2O) also supplemented with antibiotics that was grown overnight at 37°C while shaking at 180 rpm. The pre-culture was used to inoculate 2 L of NZYM contained in 2 independent baffled flasks each supplemented with 50 mg/mL carbenicillin, these culture flasks were incubated in an orbital shaker at 180 rpm and 37°C. After 2 h, the culture reached OD600 = 0.6; at this stage 500 μL of 1 mM isopropyl-β-D-1-thiogalactopyranoside (IPTG) was added and the cells continued shaking at the same speed and temperature for 4 h. The bacterial pellet was collected by centrifugation at 4,000 g and 4°C for 30 min and resuspended in 50 mL ice-cold lysis buffer (50 mM CAPS, 5 mM EDTA, 5 mM tris(2-carboxyethyl)phosphine (TCEP), 1 Roche complete inhibitor cocktail tablet at pH = 10). Cell lysis was performed on an ice-cold bath by sonication using a Sonics Vibra-Cell™ CV18 model ultrasonic processor (applying 60 power of the probe in 6 cycles of pulses, 15 s ON and 30 s OFF). The suspension was then cleared by centrifugation at 20,000 g for 45 min, and the supernatant was placed into a water bath at 100°C for 5 min. After boiling, the solution was cleared by centrifugation at 20,000 g for 45 min and the supernatant was filtered through a 0.5 μm filter before injection on a QFF HiTrap 5 mL column (GE Healthcare) for anionic exchange. Binding buffer (50 mM CAPS, 5 mM TCEP, pH = 10) was used to equilibrate the column and the supernatant was passed through the column at 2 mL/min flow using an AKTA Pure system. While the ultraviolet (UV) absorbance at 215 nm was used to monitor protein elution. Weakly bound proteins were washed out with 5 column volumes (CV) of 2% elution buffer (50 mM CAPS, 250 mM NaCl, 5 mM TCEP, pH = 10) before starting a linear gradient between 2 and 50% (using same elution buffer) without changing the flow. Fractions containing the protein of interest were identified by SDS-PAGE (detection with PageBlueTM protein staining solution, ThermoScientificTM, Cat. No. 24620) and concentrated using 20 mL spinning filters with a 3 kDa cut-off (Vivaspin 20, Sartorius AG). The concentrated sample containing ERD14-C186 and some minor contaminants was injected onto a Superdex75 16/100 HiLoad column (GE Healthcare) and the gel filtration buffer consisted of phosphate buffered saline (PBS) pH = 7.5 supplemented with 5 mM TCEP). Fractions were collected automatically based on the UV absorbance profile and SDS-PAGE allowed the identification of the protein of interest at the corresponding retention time. Selected fractions were concentrated above 150 μM for use in labeling reactions.

Alexa Fluor® 488 C5 Maleimide (green) and Alexa Fluor® 647 C2 Maleimide Alexa Fluor 647 (red) were purchased as lyophilized salts from ThermoScientificTM, Cat.No.A10254 and Cat.No.A20347, respectively. A HiTrap Desalting 5 mL column (GE Healthcare) was used to remove TCEP from the protein sample, to which either Alexa488 maleimide or Alexa Fluor 647 maleimide reagents were immediately added. The reaction mixtures (5-fold excess of fluorescent dye compared to protein concentration) were placed on a thermal bath at 60°C for 1 h. Labeling of ERD14-C186 was confirmed by SDS-PAGE through the presence of a fluorescent band with the molecular size corresponding to an ERD14 monomer. An additional gel filtration step was performed to separate the labeled protein from disulfide-linked dimers and unreacted fluorophores.

Electroporation experiments were performed on a Gene Pulser MXCellTM Electroporation System using 2 mm cuvettes (Bio-Rad Laboratories Inc., Cat No. 165-2086). A volume of 40 μL of non-modified BY-2 cells was placed inside the cuvette together with 10 μL of 250 μM ERD14-C186-Alexa488 in PBS. A broad range of conditions were tested in order to assess protein uptake and cell survivability. These conditions are summarized in Table 1 in the Results section. In every case, cells were electroporated in BY-2 medium and cargo protein ERD14-C186-Alexa488 or ERD14-C186-Alexa Fluor 647 was added in PBS. Cells were collected from the cuvette and resuspended in 100 μL of BY-2 medium. At this stage, cells were allowed to recover for different periods of time before imaging with either confocal or epifluorescence microscopy.

Cell viability was inspected by mixing electroporated cells with either lipophilic styryl dye (FMTM Thermo Fisher Scientific Inc., Cat. No. F35355) or propidium iodide (PI, Thermo Fisher Scientific Inc., Cat.No.P3566) directly on the cover slide. For cell culture monitoring and calli selection, an inverted epifluorescence microscope (ZEISS Axiovert 135M) was used. Excitation is done with an HBO 100 Hg lamp. Images were captured with a ZEISS Axiocam using the software package axiovision 4.8.2 SP2 (06-2012). Using single GFP filter and RsGFP ex 465-490 nm; em 500–520 nm. Lenses used are either plan-apochromat 20x dry na = 0.75 or plan apochromat 40x dry na = 0.95. In confocal imaging a LSM5 Exciter upright microscope using an AxioImager Z1 microscopy stand was used. Different lenses were used and are described as follows: plan-apochromat 20x dry na = 0.8, C-apochromat 63x water corrected na = 1.2, C-apochromat 40x water corrected na = 1.2. In every case imaging using dual GFP/RFP scanning was performed using software package ZEN2009.

NMR experiments (SOFAST HN-HSQC; heteronuclear single quantum coherence) (Schanda et al., 2005) were performed on a 750 MHz Bruker250 Avance spectrometer equipped with a cryogenically cooled triple resonance 1H{13C/15N} TCI probe. Samples were loaded into 3 mm Shigemi tubes and supplemented with 10% D2O. Spectra of samples in buffer were collected using 128 scans and 1,024 increments while samples in extract were collected using 256 scans and 2,048 increments. Samples recorded either for reference (in buffer) or in cell lysate were 25 μM ERD14 or ERD10 in 50 mM MES buffer; pH = 6.5. Crude Extracts (cell lysates) were obtained by sonicating 4 days old BY-2 cells on an ice-cold bath using a Sonics Vibra-CellTM CV18 model ultrasonic processor (applying 60 power of the probe in 2 cycles of pulses, 10 s ON and 20 s OFF). Cells were suspended in 50 mM MES buffer in presence of 10 mM DTT, supplemented by cOmplete protease inhibitor and PhosSTOP phosphatase inhibitor. Extracts were not centrifuged or cleared of debris. The total protein concentration was determined by the adjusted Bradford assay for NanoDropTM 2000c according to the technical note from the manufacturer.

Western blotting was performed using secondary antibodies at two different wavelengths, namely IRDye 680LT Goat anti-Mouse IgG for 680 nm and IRDye 800CW Goat Anti-Rabbit IgG for 800 nm. Imaging of developed membranes was done using dual-wavelength scanning using a LiCor Odessey® CLx. The primary antibody against ERD10, affinity purified anti-K segment antibody (Agrisera, Cat. No. AS07-206A).

It is reported that both endogenous ERD14 and ERD10 are over-expressed in planta under abiotic stress, like low temperature and drought (Hundertmark and Hincha, 2008). The quest for the structural and functional characterization of ERD14 relied on in vitro techniques, including NMR, and revealed that the protein is intrinsically disordered (Kovacs et al., 2008b; Szalainé Ágoston et al., 2011). The NMR chemical shift assignment for ERD14 is already reported (BMRB entry 26636) and further calculations supported the idea that stable secondary-structure elements are not present (Close, 1996; Szalainé Ágoston et al., 2011). The NMR assignment of ERD10 was recently made publicly available by our group (Cedeño et al., 2017; BMRB entry 26949). The availability of the chemical shift assignment for both proteins is a prerequisite when evaluating the feasibility to conduct in-cell NMR experiments (Figure 1). With such an in-cell NMR approach, it becomes possible to probe if the in vitro determined structural and dynamical properties actually reflect the cellular in vivo conformation (Selenko and Wagner, 2007). Other interesting questions relate to the identity and structural influence of binding partners, the effect of post-translational modifications on protein conformation and the ways the proteins respond structurally to cellular processes like cell differentiation. To address the high resolution conformational behavior of the ERD proteins in response to external environmental changes (e.g., abiotic stress), the development of an in-cell NMR methodology is indispensable.

Based on the already available NMR data, we opted to use the same experimental set-up, including the slightly acidic pH of 6.5 that limits the amide-proton exchange (Szalainé Ágoston et al., 2011). When we studied both the purified ERD14 and ERD10 by in vitro NMR, regardless of their size, we observed a relatively low overlap of the cross-peaks in their ¹⁵N-¹H-HSQC spectra (see Figures 2A, 3). Besides these NMR experiments in a defined buffer, we also collected¹⁵N-¹H-HSQC spectra of purified recombinant ¹⁵N-labeled ERD14 and ERD10 that were added to a cell extract of BY-2 to mimic the intracellular environment (Figures 2B, 3). Although the total protein concentration in cell extracts is far from that existing in the cytoplasm, this approach provides a reasonable and often applied approximation of cellular conditions, interaction partners and metabolites, interfering with the protein. Macromolecular concentrations closer to those prevailing in the cell can be achieved by the application of polymers, such as polethylene-glycol or dextran, which does not capture specific interactions with the protein, though. Overall, we could visualize some effects of crowding, as evidenced by line broadening for both ERD protein samples. These spectra showed little changes due to local changes of pH along the sequence as can be evidenced by the chemical shift changes experienced by histidines, which are particularly sensitive to this. It is conceivable that ERD14 and ERD10 directly interact with other cellular proteins and/or components in the crude lysates, but the data shown in this contribution do not allow us to detect such interactions (Cedeño et al., 2017). Additionally we did not observe signs in the HSQC spectra that would indicate degradation of the ¹⁵N-labeled proteins during the sample preparation and NMR data acquisition. From these experiments we concluded that the NMR data of the ERD proteins under molecular crowding conditions that resemble the plant cell interior contain sufficient information to initiate further in-cell NMR studies.

To set the stage for monitoring intracellular delivery of ERD proteins, we set-out to explore the localization of ERD proteins in plant cells. Since the exact localization of ERD14 and ERD10 inside the plant cell remains elusive and debated (Close, 1996; Puhakainen et al., 2004), we created erd10 and erd14 gene-fusions with the Red Fluorescent Protein (RFP) that were introduced into BY-2 cells using A. tumefaciens transformants. Calli that exhibited the red fluorescence of ERD14-RFP or ERD10-RFP were selected under the epifluorescent microscope and then transferred to BY-2 liquid media in order to generate stable cell lines for each protein. As clearly seen in Figures 4A,D, ERD proteins exhibit a cytosolic localization under our experimental conditions. Although our observation contrasts with previous reports (on nuclear or chloroplast association), it is in agreement with a recently published independent study (Candat et al., 2014). The same protein fusionconstructs were used for transient expression on N. benthamiana leaves together with GFP as control. As can be seen in Figures 4B,E, the fusionproteins remain in the cytosolic strands, while co-expressed GFP migrates into the nucleus (Figure 4G). ERD10 was also transiently expressed on N. benthamiana leaves and showed similar localization (Figures 4C,F). GFP co-expression controls for ERD10 are shown in Figure 4H. We found that ERD14 does not change its localization upon induced cold stress and osmotic stress (that mimics dehydration). This suggests that the protective effect of this protein (Battaglia et al., 2008; Hundertmark and Hincha, 2008) is compatible with its cytosolic localization, as simulated on leaves of N. benthamiana (Figure 5, freezing is shown in Figure 5B, osmotic stress in Figure 5C). Altogether, these results suggest that ERD14 is a good candidate for in-cell NMR and that the protein is located in the cytoplasm.

Because we could localize intracellular ERD14-RFP upon induced overexpression with success, and because of the quality of the NMR data of ERD14 under molecular crowding conditions, we decided to introduce purified and recombinantly produced 15N-labeled ERD14 inside plant cells using electroporation for in-cell NMR studies. We opted specifically for electroporation because DNA and small amounts of proteins can be delivered into plant cells by electroporation, it is possible to prepare samples containing 10⁶ cells in a matter of minutes/hours, and commercially available electroporators offer flexibility in varying the voltage and current as well as pulse timing and recovery rates (Yamano et al., 2013). These latter physical parameters determine efficient delivery of the material as well as survivability of the cells after the electric shock. Yet, our approach essentially differs from a transformation with DNA by electroporation strategy in some fundamental aspects: for transformation the main goal is to deliver a limited amount of DNA into a number of cells that must survive and get selected prior to further propagation. In contrast, for in-cell NMR, most of the cells should survive the procedure and maintain their (sub) cellular integrity, while most of the cells have to take-up a sufficiently large quantity of isotope-labeled protein material.

For this purpose, a great variety of electroporation conditions were tested to identify the mildest one with as small as possible effect on cell viability. In the first instance, we used the fluorescent derivative ERD14- C186-Alexa Fluor 647 to use microscopic visualization to verify if the protein was successfully delivered into the cells. The variables that we scouted in order to explore the delivery/survivability space, are: voltage, capacitance and pulse duration (Table 1). According to the technical notes of the electroporation device, exponential decay represents the best choice in terms of maximizing cell viability during and after the experiment. Only in a few cases we could detect successful delivery of fluorescently labeled ERD14 inside BY-2 cells (Figure 6). However, most electroporation conditions compromised cell viability in such a way that it is not compatible with any downstream in vivo experiments. As can be seen in Figure 6, both protein localization and cell morphology strongly indicate non-physiological phenomena. In general, the osmotic balance of cells appears to be heavily disrupted during electroporation (Figure 6). ERD14-C186-Alexa Fluor 647 is distributed homogeneously in the cytoplasm of some cells (Figures 6E–G, red color) showing that most of the inner architecture of the cells is disrupted. Strikingly, as also seen in Figures 6F,G, ERD14-C186-Alexa Fluor 647 is not only distributed across the cytoplasm but it can also be observed inside nuclei. There is no evidence of the association of ERD14-C186-Alexa Fluor 647 with membranes.

These adverse effects do not improve during the recovery period when membrane pores opened by the electric field are supposed to close; it seems that ERD14 can diffuse freely inside cells once pores are opened. Apparently, these set of conditions can break the tonoplast (or vacuolar membrane) whereby the cellular osmotic balance is altered in such a way that is not possible to satisfy both efficient protein delivery and cell viability. Indeed, vacuoles occupy 90–95% of the total internal volume (Marty, 1999) and the pH in the lumen reaches acidic values as low as 5.5 (Barbier-Brygoo et al., 1986). Disruption of the tonoplast is very likely the reason of the uniform distribution of ERD14-C186-Alexa Fluor 647 inside the cells without the preservation of cytosolic strands (Figures 5E–G); this is in striking contrast with the images obtained for healthy cells (Figure 4D). It must be highlighted that the internal morphology shown in this figure is not physiologically relevant as this is accompanied with FM penetration and nuclear staining (Figures 6B–D). FM does not penetrate healthy cells before electroporation (Figure 6A), therefore the conclusion is that most of the cells shown in Figures 6B–D are highly compromised, either because of disruption of vacuoles, or due to the lack of fast and efficient mechanisms repairing plasma membrane pores through cytoskeleton remodeling (Zhang et al., 2015). The observed state as shown in Figure 6 is representative of all the conditions of electroporation tested (Table 1) as there are no differences or trends observed that can be attributed to changes in voltage or intensity.

This observed vacuole disruption by electroporation also offers another interpretation concerning our in vitro crowding experiments (Figures 2, 3). These experiments did not show substantial structural changes in cell lysates as compared to a simple in vitro buffer system. Cytosolic strands are the regions (Figure 4) where both ERD14 and ERD10 are localized and where total, non-homogeneous (Luby-Phelps, 2013) protein concentration is expected to be at least 200 mg/mL (Theillet et al., 2014). The large volume of water contained in vacuoles (Marty, 1999) dilutes the cytosolic strands up to 20 times, since cell lysis disrupts the plasma membrane as well as tonoplasts. In Figures 2, 3, the crude cell lysate contains 5–7 mg/mL of protein, this low total concentration of proteins is a clear evidence of the dilution effect introduced by of vacuoles. Although the application of cell lysates/extracts is a frequently used approach to approximate or mimic intracellular effects, this might not apply for physical/structural studies where an ideal crowding experiment (such as for plant proteins) needs to be performed at high total protein concentrations like those exhibited inside cells. This offers a pivotal argument to develop a true in-cell NMR strategy.

While electroporation-mediated introduction of ERD14 did not yield the desired outcome for further in-cell NMR studies, we resorted to controlled overexpression of proteins inside plant cells as reported in the past (Hellwig et al., 2004; Ohki et al., 2008). In order to enable the production of 15N-labeled ERD proteins inside either BY-2 or A. thaliana cells in suspension, we generated transgenic lines of A. thaliana and BY-2 in which erd14 and erd10 genes were controlled by an inducible estrogen promoter. Unfortunately, gene transfection of erd14 did not yield any viable culture and therefore positive erd14 cell line could not be obtained in A. thaliana. In the case of BY-2, calli and cell suspension culture were obtained for transformants of both genes (ERD14 and 10), however, only ERD10 gets overexpressed upon the addition of estradiol, as detected by immunoblotting after 40 h of induction (Figure 7).

An inducible protein expression system can offer a major advantage over protein delivery via electroporation, because cells are constantly growing under normal culture conditions. Therefore, cell viability is much less affected during expression of ERD10 and thus cultures can grow normally during the progression of the experiment (Figure 7). However, overexpression of ERD10 in BY-2 cells imposes a set of restrictions, primarily the inherent background generated by other endogenous proteins under isotope-labeling conditions. In this scenario, ERD10 needs to be 15N labeled (at least) in order to generate a useful sample for in-cell NMR. As exemplified by the E. coli overexpression systems, translation needs to be fast and strong to obtain a good signal-to-noise ratio in an in-cell ¹H-¹⁵N-HSQC experiment (Serber et al., 2001). Liquid media for plant cultures normally contain a set of inorganic salts (micronutrients) in low concentrations and a large supply of carbon in the form of sucrose (besides vitamins and the buffer, known as Murashige-Skoog media). Vitamins like myo-inositol, thiamine and the hormone auxin determines the healthy growth of BY-2 cells, however, the mixture of vitamins and hormones can differ from one species to the other. The source of nitrogen contained in the original Murashige-Skoog medium is changed for in-cell NMR samples. In this case they were replaced by 15N-containing salts to make ERD10 visible for NMR upon estradiol-induced protein expression. The first metabolic step once ammonium and nitrate reaches the cytosol is its absorption in the form of glutamine, hence this amino acid is highly abundant in a nitrogen-rich environment (e.g., in vitro cultures) (Stitt et al., 2002; Masclaux-Daubresse et al., 2006). As shown in Figure 3 the in vitro NMR spectra of ERD10 in buffer and in a cellular extract depict the lower limit of detection of a ¹H-¹⁵N-HSQC with an acceptable signal-to-noise ratio given the low concentration and pH differences. Only qualitative information about ERD10 (fingerprint of disorder, post-translational modifications, degradation, etc.) can be extracted from these spectra (Figure 3) although lower amounts of protein would not yield easily interpretable data. These experiments show that about 12.5 μM is the lower limit of ERD10 at which NMR signals can be generated and collected with acceptable good signal-to-noise ratio, which sets our practical limit for the feasibility of the in-cell NMR experiments This was not the case in our overexpression experiment: the levels of ERD10 in BY-2 cell cultures after 48 h of induction can rise up to 2.5 μM as semi-quantitatively measured by Western Blot (Figure 7). When we compare the spectrum of a highly concentrated sample of ERD10 overlapped on the spectrum of BY-2 cells containing 2.5 μM ERD10 (Figure 8A), it is clear that sharp and intense signals can be collected from a highly concentrated sample, however, the signals coming from labeled proteins inside BY-2 are not detectable. Instead, two very intense and broad lines are detected in the region of glutamine side chains (Figure 8B), which suggests that BY-2 cells are actively metabolizing nitrate and incorporating ¹⁵N, although the translation products of other proteinaceous species including ERD10 are not detectable. This observation is in accordance with quantification of protein levels by Western blot.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.