Imaging a specific mRNA in pollen with atomic force microscopy

Department of Chemistry, Pohang Univer 790–784, Republic of Korea. E-mail: jwpark Center for Core Research Facilities, DGIST, Center for Plant Aging Research, Institute Republic of Korea. E-mail: nam@dgist.ac.kr Department of New Biology, DGIST, Daegu † Electronic supplementary informa 10.1039/c5ra00199d ‡ These authors contributed equally to th § Present address: Samsung Electronics, H Korea. Cite this: RSC Adv., 2015, 5, 18858


Introduction
Messenger RNAs (mRNAs) are localized at discrete cellular regions by several transport pathways. 1-3 mRNA localization is essential for organizing cellular architecture and function. 4 Therefore, unveiling the cellular distribution of localized mRNAs with high resolution is crucial for our understanding of fundamental cellular processes. 5,6 Several methods have been used to visualize cellular mRNAs. Fluorescence in situ hybridization (FISH) utilizes uorescently labeled oligonucleotides, proteins, or aptamers to localize mRNAs. [7][8][9][10][11] Electron microscopy, such as a scanning electron microscopy (SEM) and transmission electron microscopy (TEM), has also been employed. [12][13][14] However, ne mapping of individual mRNAs within a whole cell has yet to be achieved. 4 The atomic force microscope (AFM) has become a powerful tool for bio-imaging. Recent studies have proven the value of AFM for morphological imaging of biological surfaces and for mapping the forces in biological interactions under physiological conditions. [15][16][17][18][19][20] Because AFM has spatial resolution on the nanometer scale, AFM images provide structural information of high lateral resolution that augments the results from the other approaches. Furthermore, interaction forces between biomolecules are detected sensitively, down to a sensitivity of ten piconewtons. 21 In our previous study, a newly developed self-assembly process was employed to enhance recognition efficiency and suppress undesirable nonspecic binding during force measurement. [22][23][24][25][26] We reported imaging of Pax6 mRNA distribution in mouse brain tissue, 23 and more recently, demonstrated its capability to quantify the concentration of hepatitis C virus (HCV) RNA without labeling or amplication. 27 In the current work, we investigate the subcellular localization and distribution of AtAMT1;4 mRNA, which encodes an ammonium transporter (AMT), 28 in the sperm cells of the model plant Arabidopsis. Ammonium is the main source of nitrogen for plant growth and development. Earlier studies reported that ammonium plays a key role during nitrogen assimilation in plants, and its membrane transport is mediated primarily by the AMT family. [29][30][31][32] Recently, a pollen-specic high-affinity ammonium transporter of the plasma membrane in a pollen cell, AtAMT1;4, has been characterized. 28 Here, we utilized AFM-based mRNA mapping technology to map the AtAMT1;4 mRNA molecules within a whole pollen cell with a spatial resolution of nanometers and with singlemolecule mRNA sensitivity. Our initial expectation was that the mRNA would be mostly localized in the cytosol in the vegetative cells to provide ammonia, a nitrogen source, to the pollen cells. However, we found that the mRNA was abundant in the nuclei of sperm cells, where it is not translated.

Plant materials
Arabidopsis thaliana plant, ecotype Columbia (Col-0) was grown on soil in an environmentally controlled growth room with a 16 h photoperiod of 120 mmol photons per m 2 s À1 at 22/18 C day/ night temperature.

General information
Silicon nitride AFM probes were purchased from NanoInk, Inc (PEN-0012-03), and a probe type (S-4) of which spring constant is 16 pN nm À1 and tip radius is about 15 nm was used. A silane coupling agent N-(3-(triethoxysilyl)propyl)-O-polyethyleneoxide urethane was purchased from Gelest. All other chemicals are of reagent grade from Sigma-Aldrich. Deionized water was obtained by passing distilled water through a Barnstead E-pure 3-Module system. All water used in RNA experiments was pretreated overnight with diethylpyrocarbonate (0.05% v/v) and subsequently autoclaved. All oligonucleotides were purchased from Samchully Pharm (Korea).

Preparation of the AFM tips and substrates
A previously reported method was employed with modications for preparing the DNA-immobilized AFM probes and silicon wafers. 22-24 AFM probes and silicon wafers were treated with N-(triethoxysilylpropyl)-O-polyethylene oxide urethane (TPU) to produce an interlayer with a terminal hydroxyl group. Then, a dendron layer was introduced on the hydroxylated surface via an esterication reaction. A subsequent deprotection step generated a reactive amine group at the apex of the immobilized dendron. The amine group was treated with di(N-succinimidyl) carbonate (DSC), and then the resulting NHS (N-hydroxysuccinimide) group was used for conjugation with the oligonucleotides.

Preparation of 27-acid modied surfaces
The above silylated substrates and AFM probes were dipped for 12-24 h in a methylene chloride solution containing the 27-acid dendron (1.0 mM), a coupling agent, 1,3-dicyclohexylcarbodiimide (DCC) (29.7 mM), and 4-dimethylaminopyridine (DMAP) (2.9 mM). The 27-acid dendron (or 27-acid) used in this research was prepared by ourselves. For the preparation, an extension of the published approach 33 was employed, and 27acid was dissolved in 10 ml of dimethylformamide (DMF) before adding into methylene chloride (90 ml). Aer the reaction, the substrates were cleaned by sonication in methylene chloride, methanol, and deionized water for 3 min sequentially. The AFM probes were rinsed with methylene chloride, methanol, and water sequentially. Finally, the substrates and AFM probes were placed in a vacuum (30-40 mTorr).
Secondary structure of target mRNA and the probe design The ORF of ammonium transporter AtAMT1;4 mRNA in Arabidopsis consists of 1515 nucleotides and encodes a protein of 504 amino acids (molecular mass of 53.7 kDa). Typically, mRNA folds to form its unique secondary structure, driven by the hybridization of the self-complementary sequences. There is an enhanced hybridizing probability between the mRNA and the probe DNA if the mRNA detection region exists as a single strand. For this reason, the secondary structure of the 1515 nucleotides was simulated for analysis using an available program (http:// www.genebee.msu.su/services/rna2_reduced.html). A region of AtAMT1;4 mRNA, RNA sequences 944-977 (5 0 -AAA GAC UUA UUG AUG GGU AUU GGA AUG UAA CU GA-3 0 ), was selected for the study, and a 34-mer DNA complementary of this region was employed as the DNA probe [5 0 -NH 2 (CH 2 ) 6  The DNA corresponding to the ORF (open reading frame) of AtAMT1;4 mRNA was amplied by PCR (polymerase chain reaction) from genomic DNA of the Arabidopsis Col-0 using two specic primers, AtAMT-Nter, 5 0 -ACA ATG GCG TCG GCT CTC TCT TGC TC-3 0 and AtAMT-Cter, 5 0 -AAT CAA ACA CCT ACA TTG GGA TCA TTA-3 0 . The sequence information on Arabidopsis AtAMT1;4 DNA (At4g28700), which was possible due to the absence of introns, was obtained from a database. This 1515-nucleotide PCR product was cloned into a pGEM-T Easy vector (Promega, USA) and conrmed by sequencing. The resulting pGEM-AtAMT1;4 vector was linearized with Sal I and Aat II and used as a template to synthesize the 1515-nucleotide sense and antisense RNA of AtAMT1;4 mRNA, respectively, performing an in vitro transcription using SP6/T7 transcription kits (Roche Diagnostics, Germany). Aer transcription, the cDNA template was removed with RNase-free DNase I. The remaining RNA solution was adjusted to 0.4 M LiCl and centrifuged to pellet the precipitated RNA. The nal concentration of RNA was quantied by UV spectrometry.

Immobilization of the 1515-nucleotide cRNA on a glass slide
The mRNA was dissolved in sodium phosphate buffer (150 mM, pH 7.4) to make a solution of 0.50 mg ml À1 . Using a gel loading pipette tip, 1.0 ml of the RNA solution was placed onto a glass slide (ProbeOn Plus, Fisher Scientic, USA) and was allowed to stand at room temperature for 30 min. The drying process led to a spot of 5 mm in diameter. To immobilize the RNA on the glass surface, the slide was heated in an oven at 65 C for 30 min, and was subsequently irradiated with UV light (120 mJ) for 2 min 40 s with a UV Stratalinker (Stratagene, USA). The RNAimmobilized glass slide was then incubated in a blocking buffer [50% formamide, 10% dextran sulfate, 250 mg ml À1 yeast tRNA, 0.30 M NaCl, 20 mM Tris-HCl (pH 8.0), 5.0 mM EDTA, 10 mM sodium phosphate, 1.0% sarcosyl, 0.10% bovine serum albumin, 0.10% coll, and 0.10% polyvinylpyrollidone] at 65 C for 1 h. The slide was washed by dipping in PBS buffer ve times for 10 min each time. Finally, the slide was kept in water for a few seconds before air-drying. For the force measurements, the air-dried slide was again rehydrated in PBS buffer.

Preparation of sections of pollen cells on the glass slide using high-pressure freezing methods
Mature pollen grains from ower buds were dissected and placed into a sample carrier with 1-hexadecene and then frozen in a high-pressure freezer (BAL-TEC HPM010, Leica Microsystems). The frozen samples were transferred to a freeze substitution system at À140 C, and the temperature was raised gradually to À90 C for 48 h, and then to À60 C for 8 h with cooled ethanol. Aer dehydration with ethanol, the samples were inltrated into LR White-embedding medium (London Resign Company Ltd., UK). The samples in the embedding medium were polymerized in an oven at 55 C for 24 h. A series of 100-200 nm thick sections of pollen was prepared with an ultra AFM knife (Diatome, USA) and was placed onto glass slides (Nexterion Glass B, Schott Nexterion, Germany) coated with BIOBOND Tissue Section Adhesive (BBInternational, England). The sections were xed with distilled water, and the slides were heated in an oven at 60 C for 2 h. Finally, the slides were rinsed with PBS solution prior to AFM analysis. In this experiment, we adopted the sectioning method that has been successfully utilized for in situ detection of proteins and for EM-based detection of mRNA in a cell on the exposed surface of the section. Our sectioning method using by acrylic resins, such as LR White and Lowicryl K4 that provide the hydrophilic environment adequate for hybridization, is described in a few references including the report by Herpers et al. 14 In this experiment, we used the LR White resin. The hydrophilic surface of the LR White resin is accessible "without etching and any pretreatment" (Quoted from the LR White manufacturer manuals, London Resin Company Ltd., UK).

AFM force measurements
All force measurements were performed with a NanoWizard AFM (JPK Instruments, Germany). The spring constant of each AFM tip was calibrated in solution before each experiment with the built-in program. The spring constant of the cantilevers was between 14 and 25 pN nm À1 . All measurements were carried out in fresh DEPC (diethylpyrocarbonate)-treated PBS buffer (pH 7.4) at room temperature. All force measurements were recorded with a tip velocity of 0.54 mm s À1 . Force measurements of model systems were as follows: to measure the mean force values, the force-distance curves were recorded at least 100 times at one position on a substrate, and more than ve spots were examined in each experiment. Force measurements of a sectioned cell were as follows: force mapping experiments were performed immediately aer morphological scanning on the sectioned cell surface. Topographic scanned area was divided into 50 Â 50 (2500) pixels. Force maps were obtained by processing the force values recorded during the raster-scanning on areas. Typically force-distance curves were recorded ve times at each pixel, and the presented force value of each pixel is the average of the unbinding-force values.

AFM imaging on sectioned pollen surface
All AFM images were obtained with a NanoWizard AFM (JPK Instruments, Germany) using intermittent contact mode in a uid environment [PBS (phosphate buffered saline) buffer, pH 7.4] with 1.0 Hz tip velocity. The scan line was 512.

Transmission electron microscopy analysis
To observe the components of pollen, mature pollen grains were dissected and xed with the high-pressurized liquid nitrogen system in a high-pressure freezer (BAL-TEC HPM010, Leica Microsystems). The pollen samples were post-xed with 2% osmium tetroxide (OsO 4 ) and were dehydrated with a graded series of ethanol. Aer dehydration, the pollen samples were inltrated with propylene oxide twice for 30 min each, and then inltrated into EMbed-812 embedding medium (Electron Microscopy Science, USA). The samples in the embedding medium were polymerized at 60 C for 30 h. Pollen sections (70-80 nm thickness) were stained with 2% uranium acetate followed by 2% lead citrate and visualized using a JEOL JEM-1011 transmission electron microscope (JEOL, Tokyo, Japan). Cry-oxation of Arabidopsis pollen was performed as previously published 34 with some modications. Briey, the frozen samples were transferred to a freeze substitution system at À140 C, and the temperature was raised gradually to À90 C for 48 h and then to À60 C for 8 h with cooled acetone.

Results and discussion
Force mapping the AtAMT1;4 mRNAs in model systems Model system I: interaction between the DNA probe and the 34-mer complementary RNA. To detect a selected target region of AtAMT1;4 mRNA, a 34-mer DNA oligonucleotide that was complementary to the mRNA sequence between nucleotides 944 and 977 was used as a probe (Fig. S1; † and for details, see Methods). In order to nd sequences that allow facile interaction with the probe DNA, the secondary structure was simulated, and regions of single strand were located. One of the candidates in the middle part of mRNA was chosen as the detection part. For the enhanced specicity, the sequence of 34 bases was utilized instead of the previously employed length, 18 bases. 27 The interaction force between the DNA probe immobilized on the AFM tip and the 34-mer target RNA immobilized on a substrate is shown in Fig. 1A. The DNA probe complementary to nucleotides 944-977 was conjugated with an AFM probe treated with the TPU and 27-acid dendron, and the 34-mer RNA was conjugated on a silicon wafer treated with the silane reagent and the dendron. The force-distance curves of the interaction between the DNA probe and the target RNA were obtained by the approach and retraction method. The measurement was repeated 200 times with a tip velocity of 0.54 mm s À1 , and 148 specic curves were recorded. The resulting force-distance curves showed a single unbinding (adhesive) pattern during the retraction process (Fig. 1B). The total unbinding-force curves were used to generate a force histogram (Fig. 1C). Gaussian tting of the histogram yielded the most probable force of 44 AE 1 pN (n ¼ 148 specic curves/200 measurements; s ¼ 3) for the unbinding event. To verify the specicity, the interactions between the DNA probe and noncomplementary target RNAs [5 0 -NH 2 (CH 2 ) 6 -UC AGU UAC AUU CCA AUA CCC AUC AAU AAG UCU UU-3 0 ] were measured. The unbinding frequency was dramatically reduced to 7%, and the mean unbinding-force (28 pN) was signicantly smaller than the specic interaction mean unbinding force (Fig. 1D).
Model system II: interaction between the DNA probe and the 1515-nucleotide AtAMT1;4 mRNA. The interaction between the 34-mer DNA probe on the dendron-modied AFM tip and cRNA on the slide was measured in the same way as for model system I ( Fig. 2A). Force-distance curves during the retraction process showed mostly a single event (Fig. 2B), and multiple unbinding events were rare. For curves with multiple events, the force value corresponding to the last rupture event was calculated, and the value was incorporated into the histogram. The most probable unbinding-force value was 41 AE 1 pN (n ¼ 666/900; s ¼ 6) (Fig. 2C). To conrm the specicity, the interaction between the probe DNA and antisense mRNA was measured. The unbinding frequency was dramatically reduced to 7%, and the mean unbinding-force (28 pN) was signicantly smaller than the specic interaction mean unbinding force (Fig. 2D).

Force mapping the AtAMT1;4 mRNAs on a sectioned pollen
The most probable unbinding-force value in the above model systems was 44 AE 1 and 41 AE 1 pN, respectively, and the measurements showed characteristic stretching distances (the most probable distance of 7 nm and 11 nm, respectively). The unbinding-force was also examined using a model system exposing the RNA on a cross-sectioned surface of an embedded resin, which simulates conditions that are similar to the sectioned pollen surface (ESI, Fig. S2 †).
To investigate the interaction between the 1515-nucleotide AtAMT1;4 mRNA and the DNA probe on the sectioned pollen  surface, 100 force-distance curves were recorded. Fig. S3A † is the force histogram from the specic force curves recorded at a certain position. For curves with multiple events, the force value calculated from the last rupture event was incorporated into the force histogram. Gaussian tting gave the most probable unbinding-force of 39 AE 1 pN (n ¼ 50/100; s ¼ 3). The percentage of no event was 50%. Gaussian tting provided the most probable unbinding-distance value of 16 AE 1 nm (n ¼ 50/ 100; s ¼ 2) (Fig. S3B †). Fig. S3C and D † show typical force-distance curves recorded in this experiment. The mapping was performed by measuring the specic unbinding-force between the DNA probe and the complementary part of AtAMT1;4 mRNA (Fig. 3A). The AFM topographic image of the sectioned surface showed the subcellular structure of pollen (Fig. 3C), and the area was divided into 50 Â 50 pixels (i.e., 400 nm interval) for the force mapping. Although hundreds measurements were required at each pixel for high accuracy, a compromise strategy recording ve force-distance curves at each pixel was taken so that the entire area could be measured in six hours. Aer the force measurements were complete, an AFM topographic image was obtained again to conrm that the lateral dri was not signicant. The mean value of the measured unbinding-force and stretching distance, and the probability of obtaining the specic curves, were used as the criteria for the location of mRNAs in the map.
To set an appropriate range of the specic force value, a histogram of the force value collected from 2500 pixels (12 500 measurements) was obtained (Fig. 3B). A total of 878 specic curves were recorded (most probable force value ¼ 38 pN, standard deviation of the distribution ¼ 7.7), and Gaussian tting of the mean value from 373 pixels gave the most probable value of 38 pN and s value of 7 (n ¼ 373). The unbindingforce value of AtAMT1;4 mRNA on the sectioned cell was a few pN smaller than those of the model systems. This reduction had been observed in previous studies, and the difference was believed to be associated with plasticity of the sample surface. 35 For mapping purposes, pixels were categorized into ve groups. A pixel showing no event was grouped into 'no event', and was colored black. A pixel showing one or two nonspecic events was also classied as 'no event'. A total of 2112 out of 2500 pixels belonged to the 'no event' group. A total of 15 pixels showed a linear unbinding prole three to ve times and were categorized into the 'nonspecic' group and colored black. The remaining 373 pixels (showing at least one specic event and no more than two nonspecic events) were classied into the 'specic' group. Pixels in the 'specic' group with mean force values between 31 and 45 pN (259 pixels, AE1s) were colored yellow, pixels with mean force values between 46 and 60 pN (55 pixels) were colored light yellow, and pixels with mean force values between 16 and 30 pN (55 pixels) were colored dark grey. Other pixels with mean force values greater than 60 pN were colored dark navy (4 pixels). The most probable probability of obtaining the specic curves in the yellow pixels was 30% and the most probable stretching distance was 13 AE 2 nm. Although there was a high probability to miss RNAs exposed to the sectioned surface, as the force was measured at 400 nm interval (the lateral displacement of the sample with respect to the AFM tip), the map revealed the approximate distribution of mRNA on a sectioned pollen cell. For example, the areas where black pixels are abundant, notably some areas in the center, must represent domains in which the mRNA is very scarce (Fig. 3D). A few specic pixels were found outside of the sectioned cell, which may be due to contamination during sample preparation or experimental uncertainty. Whereas the background level was not subtracted from each force image, the background level may be useful as a baseline for comparison for other maps obtained in this study.
To understand the hydrodynamic behavior of mRNA exposed on the surface, the curves were obtained at 5.0 nm interval (Fig. 3E-I). 36 Given the complexity of mRNA itself and the variations in physical status (i.e., orientation, freedom to move in the medium, and distortion during the sectioning process), a rather broad range of hydrodynamic radii is expected. The high-resolution map showed clusters with diameters ranging from 25 to 40 nm. The size of an isolated cluster was 2530 nm (Fig. 3F and G); the most probable force value and probability of obtaining the specic curves within 16 pixels was 43 AE 1 pN and 49%, respectively. Sometimes, a larger cluster was found ( Fig. 3H and I), and it is believed that two clusters were close to each other to make the larger cluster, and the probability was more or less similar to the one described above. Therefore, recording the map at 20 nm intervals should result in a low probability of missing RNA that was accessible to the probe if the curves were recorded ve times at each pixel. To detect every mRNA exposed on the surface, it may be necessary to measure the force with an interval of smaller than the hydrodynamic radius, such as 10 nm. Generally, a resolution of 20 or 60 nm was employed in this study due to practical constraints, such as the size of the pollen cell, the subcellular structure, and the limited time window to avoid degradation of mRNA, and lateral dri of AFM. Nevertheless, maps of such resolution are expected to show the RNA distribution pattern in most studies.

Force mapping of specic domains
For accurate high-resolution mapping of mRNA on pollen sections with subcellular components, it is very important that the endogenous structures of all subcellular components were preserved during the cutting and xation process. Therefore, pollen cell sections were prepared by high-pressure freezing methods (for details, see Methods). As shown in Fig. 4A, the sperm cell nucleus and cytoplasm maintain their structures in an AFM topographic image, which is further veried by TEM analysis. These cell sections were employed for high-resolution mapping of subcellular components. An interval of 60 nm was employed to examine the whole sperm cell (Fig. 4B, right two images), and the 20 nm interval (Fig. 4C, right two images) provided greater detail for visualizing the nuclear membrane. Both maps showed that the mRNA was highly expressed inside the sperm nucleus, and was less abundant in the cytoplasm. There were 208 yellow pixels within the sperm nucleus (71 per 1.0 mm 2 ), whereas there were 41 yellow pixels in the cytoplasm of the sperm cell (16 per 1.0 mm 2 ). Thus, the relative surface concentration of the mRNA in the nucleus is approximately four times higher than in the cytoplasm in the sperm cell.
Control experiments were performed and conrmed the validity of these results. The probability of observing a specic interaction was dramatically reduced when the surface was blocked with antisense mRNA before AFM imaging and when the surface was treated with RNase ( Fig. S4 †). The antisense blocking and RNase treatment reduced the pixel number of specic event from 27% to 4% and 8%, respectively, of the sample measured for a same area. The most probable force value (24 pN for blocking with the antisense RNA and 25 pN for treating with RNase) was lower than the standard force value (38 pN). Additionally, the probability of a specic interaction was very low in petal cell sections (Fig. 4D). As shown in the force map in the gure, the probability of the unbinding event was insignicant even without the treatment of RNase or the antisense RNA.

Conclusions
Compared to the conventional mapping technologies such as confocal microscopy, the method we presented shows superior lateral resolution due to the inherent characteristics of AFM. Furthermore, this AFM method does not require any specic labeling procedure. Due to the high sensitivity of AFM force mapping, this method allows localization of single mRNAs, not requiring any amplication step. Detecting local distribution of RNA is becoming one of the key issues in understanding cellular function of RNA. Although we demonstrated the utility of this method in plant pollen cells, we argue that this method should be facilely applicable to most cells, provided that a high quality section can be achieved. We believe that direct mechanical measurement of individual mRNAs would provide much reliable and sensitive detection of mRNA than the hybridization based detection, as the former conrms the identity of an analyte through valuable observables (force value, stretching distance, and cluster radius) and the latter lacks such virtues and is inherently prone to background noise. Nevertheless, inuence of the secondary structure of chosen RNA sequence should be studied further, because the structure is one of the key factors determining the kinetic and thermodynamic behaviors of the interaction between the sequence and the probe DNA on the AFM probe. Because ribosomes associated with the RNA may deter the above interaction, impact of such proteins should be unveiled. While we demonstrated application of this method for an mRNA, we believe that this method can be applied for imaging other non-coding cellular RNAs. This method thus should provide a new leap in RNA biology with facile and efficient RNA imaging capability at the intracellular level and open a new window to understand the function and behavior of RNA molecules for organizing cellular architecture and function.