Robotic assisted generation of 2'-deoxy-2'-fluoro-modifed RNA aptamers - High performance enabling strategies in aptamer selection.

Aptamer selection is a laborious procedure, requiring expertise and significant resources. These characteristics limit the accessibility of researchers to these molecular tools. We describe a selection procedure, making use of a robotic system that allows the fully automated selection of RNA and 2'deoxy-2'-fluoro pyrimidine RNA aptamers. The platform offers a rapid access to aptamers for basic research and development, therefore opening the path to aptamer-based systemic analysis of proteomes in biological settings.

Introduction

Aptamers are short single chained nucleic acids, selected from gigantic libraries predominantly by in vitro procedures [1], [2]. Literature searches reveal an increasing demand on this compound class for research and development, e.g., as capturing ligands in diagnostic and sensing applications or as inhibitors in biomedical settings [3]. This demand is faced by a limited number of laboratories specialized in the identification and characterization of aptamers and the lack of sophisticated automation of the selection process. Because of this, the provision of aptamers for basic sciences, e.g., as inhibitor to validate target function [4] cannot take pace with the needs of other ‘omics’ disciplines and the demands of state-of-the-art life science research [5], [6]. In a seminal publication, Ellington and co-workers described an automated workstation able to conduct up to six consecutive selection cycles [7], [8]. Besides this success, other semi-automated platforms, which still includes manual selection rounds or manual assessment of PCR and RT-PCR performance have been described [9], [10]. The aptamer selection process consists of several steps, including incubation, separation, washing, recovery, amplification, and depending on the nature of the nucleic acid library used, a single strand generation step (in case of DNA) or in vitro transcription step (in case of RNA). Thus, an automated procedure needs to fine tune and balance the efficiency of each step with one another. This adaptation is challenging and certainly requires compromises to be met. An automated selection process that is capable of performing up to twelve consecutive selection cycles (which for most targets is sufficient to gain enrichment), will certainly help to overcome limitations in regard of time and costs of the aptamer generation process as well as throughput and accessibility to aptamers. Automation also offers a reproducible setting based on standardized procedures, whereas these come along with limitations on their own, e.g., cycle to cycle variations of selection stringency as possible in manual selection formats.

Here we describe a robotic assisted selection procedure, which performs up to 12 consecutive selection cycles capable of using up to 8 target proteins simultaneously. We developed a protocol that allows the automated generation of RNA and 2′-deoxy-2′-fluore pyrimidine modified RNA aptamers, without manual interference. This platform will speed up the aptamer generation process and opens the path towards rapid aptamer generation for enabling strategies and the systemic analysis of proteins. We envision the platform fueling an ‘aptanomics’ approach, in which aptamers will be rapidly provided for target proteins and subsequent validation in biological systems [11].

Configuration of the robotic selection platform

The robotic system is composed of various individual automated laboratory positioners (ALP), including a setup of different machines that are converged yielding a unique robotic set up.

We built the robotic platform using a Biomek NXP workstation, which executes all liquid handling steps. It is equipped with a SPAN 8 pipetting model enabling the operation of up to 8 samples simultaneously and a series of 12 selection cycles without manual interference. The automated selection procedure uses a 96-well microtiter plate system for executing the incubation, separation, reaction, and storage steps. 2 3D ALPs are integrated on the deck, steered by a compressed air system that enables tilting in x/y axis including a pivoting and knocking feature (Fig. 1a). We also implemented 4 ALPs with temperature control (10 °C–70 °C) for incubation and storage of samples in the microtiter plates (Fig. 1b). The deck for enzyme handling contains a freezing-position controlled by an external cryostat for lower temperatures (−20 °C, Fig. 1b), including a specially designed lid that provides the microtiter plate with a permanent buffer of dry air to prevent frosting. The working temperature of the different positions on the Biomek NXP varies from −20 °C for enzymes, 4 °C for beads and reaction mixes and 37 °C for incubation steps. For the separation step, a magnetic ALP on position ALP4 and a vacuum station on position holder_1 are included (Fig. 1b). For disposal, a waste position is defined for used labware (Fig. 1b). We integrated a microplate hotel on the platform as a repository for the required labware. For PCR, a thermocycler is embedded in the system adjacent to the Biomek NXP. During all steps performed by the PCR cycler, the microtiter plate is sealed by a special reusable PCR-lid for automation to prevent evaporation. Due to the programming, the PCR-lid can be used for one selection of 12 rounds without any risk of cross contamination.

Fig. 1

The robotic assisted selection platform. (a) Picture of the workstation and (b) its schematic work surface.

A magnetic trail system and gripper connect the deck through position ALP6 (Fig. 1) with all other devices (hotel, PCR cycler). All devices are controlled by the SAMI EX software and all pipetting steps are programmed through the Biomek software and integrated into the SAMI EX.

The automated selection procedure

Before starting the fully automated SELEX process, all required labware and stationary items are placed on the platform at the respective position. During the operating process, the platform assembles itself with all labware that is hold available in the hotel. The system uses filter tips, which are used for all pipetting steps, to decrease contamination risks. In addition, all stationary microtiter plates are sealed manually with a pre-cut pierceable foil.

To minimize the loss of pipetting accuracy during the selection procedure, the SPAN 8 performs a calibration procedure during the initialization process and after every selection cycle. Initially, the samples, i.e. target molecule on magnetic beads and starting library are placed on the platform for incubation under temperature control and gentle shaking. For homogeneity, the SPAN 8 pipets up and down every 5 min. After the incubation step, the samples are transferred to the magnetic rack to separate the magnetic target-beads and remove the unbound RNA species [12]. Subsequently, the beads are washed with several aliquots of buffer, whereupon the quantity of washing cycles can be adapted for each selection cycle and depending on the target molecule’s nature. After washing, the recovery of bound RNA species is achieved by adding double distilled H2O and heating at 80 °C for 5 min in the PCR cycler. The supernatant is then subjected to reverse transcription and PCR amplification.

After amplification, the dsDNA is transcribed into RNA, whereas an aliquot of the dsDNA product is stored in a stationary sealed 96-well storage plate at 4 °C. After transcription, the resulting RNA is incubated with the target-beads to initialize the second selection cycle.

After 8–12 selection cycles (28 h–42 h), the dsDNA storage plate is removed manually and stored at −20 °C until further processed. All other stationary labware gets detached and discarded.

We implemented 2 manually performed control steps aside the automation to validate the automated process. First, the dsDNA products of each selection cycle are analyzed by agarose gel electrophoresis. Second, interaction analysis of the RNA from the various selection cycles is done, e.g., using radioactively labeled RNA for a filter-retention assay. Typically, we compare the binding properties of the RNA obtained after selection cycle 1 with the RNA from the final selection cycle, e.g., selection cycles 8–12. In this way, we get a comprehensive view on the performance of the platform.

Exemplary results

To validate the platform, we performed a series of consecutive automated selections targeting the mitogen-activated protein (MAP) kinase extracellular signal-regulated kinase 2 (Erk2) [4]. Erk2 was used as in previous selection experiments we made good experience and found it represents an ‘aptamerogenic’ target [13]. We performed 10 consecutive selection experiments, with variations of the total amount of iterative selection cycles (i.e., 8, 10, or 12 selection cycles). Two of these selection experiments failed, due to a platform error that is apparent in an additional shorter PCR fragment (Fig. 2a). Due to the complexity of the system and processing of the selection without any assistance by a technician or video control, a clear determination of the error is not feasible. Referring to the existing PCR product in all selection rounds, a mechanical malfunction of the system can be excluded. Possible explanations might be manual mistakes in the preparation steps before automation, since this plays a decisive role for the biology later on the Biomek NXP. Furthermore, minor pipetting failures in the automation, regarding a change of volume due to small air bubbles in microtiter plates, or the air gap pipetting behavior of the Span 8 cannot be completely ruled out. The other eight selection experiments were conducted successfully. The analysis of the PCR products showed a dsDNA amplificate throughout all selection cycles (Fig. 2b). Moreover, the interaction analysis revealed an enhanced binding of all eight RNA libraries from the last selection cycles when compared to the starting library (Fig. 2c). The amount of Erk2-bound RNA however varies among the different selection experiments between 38% and 28% (Fig. 2c). These data show that our fully automated platform is not only being able to execute successful selections of RNA libraries, but also reveals a reproducibility with a coefficient of variation of 16% (Fig. 2d).

Fig. 2

Results of consecutive selections of RNA aptamers binding to Erk2. (a) Agarose gel of an unsuccessful, in terms of gaining enhanced binding of the resultant RNA libraries to Erk2 (see c), selection experiments. The gel analysis reveals the dsDNA corresponding to the library C2 at 107 bp and additional shorter dsDNA molecules at ∼80 bp; the numbers correspond to the selection cycles, as reference an Ultra Low Range DNA Ladder (ULR) was applied. (b) Agarose gel of a successful, in terms of gaining enhanced binding of the resultant RNA libraries to Erk2 (see c), selection experiment. The gel analysis reveals the C2 band at 107 bp, whereas shorter fragments are less pronounced compared to (a). The numbers correspond to the selection cycles, as reference an Ultra Low Range DNA Ladder (ULR) was applied. (c) Eight selection experiments (differently shaded grey bars 1–8) yielded RNA libraries with enhanced interaction with Erk2 [1 µM, +] and no binding in the absence of Erk2 (−). Note, the order of the selections as shown do not correspond with the order of execution. Two selection experiments (9,10) do not show enhanced binding to Erk2 [1 µM, +]. The C2 starting library (black bar, neg. ctrl.) was used as a negative control. (d) Statistical analysis showing the reproducibility of the robotic assisted selection experiments in the presence or absence of Erk2 (%cv: unsuccessful SELEX = 15.7%; %cv successful SELEX = 15.9%). The C2 starting library (neg. ctrl.) was used as a negative control.

Having established the automated protocol for obtaining RNA aptamers, we next set out to adapt towards the selection of 2′-deoxy-2′-fluoro pyrimidine (2′fRNA) RNA aptamers. We replaced the nucleotide triphosphates UTP and CTP by the 2′-deoxy-2′-fluoro modified variants and exchanged the T7 RNA polymerase by the mutant (Y639F) [14], [15]. This setup was then employed to perform an automated selection experiment, again using Erk2 as the target protein. The analysis of the PCR products after eight selection cycles revealed a regularly amplified dsDNA among the selection cycles (Fig. 3a). Filter retention analysis revealed significant binding of the enriched 2′fRNA library when compared to the starting library (Fig. 3b). We subsequently analyzed the corresponding DNA libraries from the starting library and selection cycles 5 and 8 by next generation sequencing (NGS) [16], [17]. These analyses revealed an equal distribution of the four nucleotides within the random region of the starting library (Fig. 3c), but a strong shift of each nucleotide’s frequency in the library from selection cycle 8 (Fig. 3d). The total amount of unique sequences decreased along the selection experiment from ∼91% in the starting library to below 1.5% in the library from selection cycle 8 (Fig. 3e), while a total number of reads between 2 and 8 million was obtained (Fig. 3f). We chose 3 of the most abundant sequences (Table 1) for further analysis. Affinity determination revealed kD-values between 245 and 723 nM for all three 2′fRNA aptamers binding to Erk2 (Fig. 4a). Specificity determination discloses binding for all three 2′fRNA aptamers to Erk2 and strongly decreased binding to Mek1 wildtype and the active mutant of Mek1, referred to as Mek1 G7B (Fig. 4b).

Fig. 3

Results of the robotic assisted selection of 2′fRNA aptamers binding to Erk2 and NGS analysis. (a) The gel analysis of the obtained dsDNA from the different selection cycles of the 2′fRNA aptamer selection shows a constant PCR product at 107 bp. Note, some shorter fragments are visible in the first selection cycle, but these are vanishing during the progression of the selection. The numbers correspond to the selection cycles, as reference an Ultra Low Range DNA Ladder (ULR) was applied. (b) The 2′fRNA libraries obtained from the selection cycles 1 and 8 were analyzed by filter retention analysis and reveal an enrichment of binding species (n = 2, mean ± SD). (c) Nucleotide distribution at the different positions of the random region in the C2 starting library. (d) Nucleotide distribution at the different positions of the random region of the library from the selection cycle 8. (e) Frequency of unique sequences in the C2 starting library (SL = 91.4%), selection cycle five (5 = 21.5%) and selection cycle eight (8 = 1.5%). (f) Number of reads in the NGS analysis in the starting library (SL = 2 × 106), selection cycle five (5 = 8 × 106) and selection cycle eight (8 = 7 × 106).

Table 1

Showing the most 3 abundant sequences relying to their motifs in relation to their amount of copies and frequency of the sequence-family.

ID	Sequence of random region	Copy numbers		Frequency [%]		Kd [nM]
		Pool	Round 8	Pool	Round 8
R2F 1	CCTCCAGGGATGGGTCTACGTGTATTACGGGGGTATTTGG	136	1,436,246	0.01	19.14	245
R2F 2	CTGGCACTTTACATCCGTATTGGTTGTGTCGGTTCGATTG	134	861,914	0.01	11.49	434
R2F 12_1	GGTAAAGATACGTCTACGAGTAATGAAGTCGGCTCGGGGG	592	2,416,670	0.03	32.21	723

Fig. 4

Dissociation constants and specificity for the obtained 2′fRNA aptamers. (a) A filter retention analysis of the three most abundant sequences reveals kD values of 245 nM (R2F 1, black curve), 434 nM (R2F 2, blue curve), and 723 nM (R2F 12_1, red curve). (b) A filter retention analysis of the most abundant three sequences show binding to Erk2 [1 µM] and strongly decreased binding to Mek1 wildtype or its constitutive active mutant G7B. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Comparison automation and manual SELEX

The fully automated selection process has several advantages over manually driven processes. First, performing a selection of RNA aptamers for one target manually is a process of approximately two weeks, considering only the selection without any evaluation of the obtained data, i.e. interaction analysis. Compared to that, the automated process performs a selection of 12 rounds in less than two days, which reveals a saving in time of ∼80% (Fig. 5). Regarding the potential of performing up to eight selections simultaneously, this saving is increased significantly. Second, the random and operational error is decreased, as the programming and the robotic system do not allow changes during the process. The relative error in all steps of the process is mainly driven by the pipetting error of the SPAN-8 system, which is comparable to the regular pipetting error of every other conventional pipette. Major errors as they may occur in the manual procedures, in particular when multiple targets are handled simultaneously, can be excluded by the robotic system. Third, there is nearly no risk of incorporation of hazardous compounds during the process, since there is no direct contact between operator and compound during the automated selection process. Fourth, the time-to-aptamer is significantly shortened and the process is applicable to many researchers without the need to get trained in selection experiments.

Fig. 5

Comparison of the processing time of the manual and robotic assisted selection. The automation takes 1.78 days for 12 cycles of selection, while a manual selection takes on average 10.46 days ± 1.76 days (n = 13, %cv = 27.21) for 12 cycles of selection.

Automation, however, is also associated with two limitations. First, processing of a selection experiment manually enables the possibility to monitor every single selection cycle and to visualize potential failure. A missing PCR-product or additional PCR bands [18] indicate an error during the process and lead to different strategies to prevent more damage. Since the repetition of the selection cycle in a manual process could solve some of these problems, the automated selection will simply continue, which results in a loss of material. Second, depending on the SELEX strategy, a broad range of laboratory equipment can be used to achieve a successful manual selection. The robotic system is limiting the variability of applicable selection strategies.

In conclusion, we describe an approach for the rapid generation of aptamers using a robotic based approach. Our set up allows up to 12 consecutive selection cycles to be conducted without manual interference, a yet unmet performance. We not only show the platform to enrich for RNA aptamers but also being compatible for the identification of 2′-deoxy-2′-fluoro modified RNA aptamers, exemplified using Erk2 as target protein. The protocol outperforms the time requirements of manual selection procedures 6-fold considering single target usage. The modular design of the platform will allow also the selection of DNA aptamers and establishing other selection formats, e.g., capture-SELEX. Currently, the described system is restricted by performing the selection process, but additional configurations and amendments to the platform will allow the implementation of interaction analysis and sequencing techniques [19]. Thereby, the entire process will be adaptable to high-throughput standards, speeding up the time-to-aptamer significantly.

Material and methods

Coupling of Erk2 to Dynabeads His-Tag isolation & pulldown

Dynabeads His-Tag Isolation & Pulldown (ThermoFisher Scientific) were used for immobilization of Erk2. Therefor 100 µL of bead solution (40 mg beads/mL) were used for 9.6 pmol of Erk2, prepared in 1 mL binding/wash buffer (50 mM Sodium-Phosphate, pH 8.0, 300 mM NaCl, 0.01% Tween®-20) according to the manufacturers protocol. Prior to coupling, the provided buffer of the bead solution was discarded using a DynaMag™-2 Magnet (ThermoFisher Scientific) for separation. The coupling reaction was incubated on 4 °C for 30 min, involving a Tube Revolver Model D-6050 (neoLab) rotating at a speed of 200 rpm, followed by 3 washing steps with 1 mL binding/wash buffer according to the manufacturers protocol. Beforehand resuspending the Erk2 beads in 1 mL of storing buffer (1.25× PBS; 171.25 mM NaCl (Fisher Scientific), 3.38 mM KCl (Roth), 12.5 mM Na2HPO4 (Roth), 2.2 mM KH2PO4 (Roth), pH 7.4; Albumin (BSA) Fraction V (pH 7.0) 1 mg/mL (AppliChem)), the sample was washed additionally once more with storing buffer according to the manufacturers protocol.

Automated selection

C2 DNA library (5′ – GGG AGA GGA GGG AGA UAG AUA UCA A – N40 – UUC GUG GAU GCC ACA GGA C − 3′) and primers were synthesised by Ella Biotech GmbH (Munich, Germany). C2 DNA library was transcribed in vitro by using T7-RNA polymerase (NEB) and RNasin Inhibitor (Promega) in transcription master mix (40 mM Tris-HCl [pH 7.9], 15 mM MgCl2 (Roth), 5 mM DTT (AppliChem), 2.5 mM NTP (Jena Bioscience) each). For in vitro transcription of C2 2′f RNA library, 2′f UTP and 2′f CTP of METKINEN chemistry (Kuopio, Finland) and the T7-RNA polymerase mutant Y639F were used [11]. Both libraries were amplified by PCR using the following primers: forward primer including the T7-promoter sequence (C2 fw) 5′ – AATTCTAATACGACTCACTATAGGGAGAGGAGGGAGATAGATATCAA – 3′ and reverse primer (C2 rv) 5′ – GTCCTGTGGCATCCACGAAA – 3′ in PCR master mix (1× colorless GoTaq® Flexi Buffer (Promega), 2 mM DTT (AppliChem), 1.5 mM MgCl2 (Roth), 0.3 mM dNTP (Genaxxon) each). Reverse transcription was performed by using M-MLV Reverse Transcriptase (Promega). RT-PCR reaction was performed by using GoTaq® G2 Flexi DNA Polymerase (Promega) and 1 µM of both C2 fw and C2 rv primers in a total reaction volume of 100 µL with the following cycling program (10 min 55 °C; 60 s 95 °C, 60 s 56 °C, 60 s 72 °C; 3 min 72 °C; hold 10 °C) in a TRobot thermal cycler (Biometra) for 19 PCR cycles in the first 4 selection rounds and 16 PCR cycles in every selection round after. Reverse transcription and PCR were performed in a one-time reaction, first pipetting RT-PCR-Mix and template together for a 5 min incubation at 65 °C. (Adding) After incubation, 1 µL M-MLV (Promega) and 1 µL GoTaq® G2 Flexi DNA Polymerase (Promega) were added per reaction (after incubation) to start (starting) the RT-PCR (there was no verb). In all PCR (cycler) steps an arched auto-sealing lid for PCR plates (Bio-Rad) was used.

Automated selection was done using a Biomek NXP (Beckman Coulter). 0.5 nmol of C2 RNA library was pipetted to Erk2 immobilized on Dynabeads His-Tag Isolation & Pulldown (ThermoFisher Scientific) and incubated for 30 min at 37 °C while shaking at a speed of 700 rpm; additional pipetting up and down every 5 min. Prior to pipetting, the library was prepared with MgCl2 (Roth) and phosphate-buffered saline (PBS; 137 mM NaCl (Fisher Scientific), 2.7 mM KCl (Roth), 10 mM Na2HPO4 (Roth), 1.76 mM KH2PO4 (Roth), pH 7.4) for a final concentration in the selection of 3 mM MgCl2 and PBS. After incubation the samples were washed 2 times with wash buffer (3 mM MgCl2, PBS). The number of washing steps were increased every round by 2 washes, up to a total of 8 washes. An elution step was performed with double distilled H2O (TKA Wasseraufbereitungssysteme GmbH) for 5 min at 80 °C in a TRobot thermal cycler. Prior to RT-PCR, the samples were incubated for 5 min at 65 °C in a TRobot thermal cycler. After RT-PCR the in vitro transcription (was performed) followed, containing 10 µL of PCR product in a total in vitro transcription volume of 100 µL. Transcription was performed by the Biomek NXP for 30 min (RNA selection) and 45 min (2′fRNA selection) at 37 °C. 20 µL of the transcribed RNA were directly transferred to 80 µL of beads suspension for the next selection round.

Agarose gel analysis

A 4% agarose Gel with Agarose LE (Genaxxon) was performed and pre-stained with ethidium bromide (Roth). 5 µL of dsDNA mixed with 1 µL 6x DNA Loading Dye (Thermo Scientific) and 4 µL of GeneRuler Ultra Low Range DNA Ladder (Thermo Scientific) were loaded on the gel and kept running for 15 min at 150 V. Visualization of the DNA was performed with a Genoplex system (VWR).

Filter retention analysis

The corresponding ssDNA molecules of R2F 1, R2F 2 and R2F 12_1 were synthesised and PCR amplified to yield dsDNA. The dsDNA was transcribed in vitro, in the presence of 10 µCi α-32P-GTP (PerkinElmer). After purification, the radioactively labeled RNA was incubated with protein mix in 3 mM MgCl2 and PBS for 30 min at 37 °C. Filtration was done with a Nitrocellulose Blotting Membrane (Amersham Protran 0.45 µm NC), which was prior buffered for 30 min in cathode buffer (27 mM TRIS (Roth), 40 mM 6-Aminohexanoic acid (Alfa Aesar), pH 9.4). After filtration the membrane was washed 3 times with washing buffer (3 mM MgCl2, PBS). The Nitrocellulose Blotting Membrane was analyzed with a phosphor screen by a Phosphorimager (FujiFilm FLA3000) after a minimum of 5 h of radiation. The analysis was performed with the software Advanced Image Data Analyzer (AIDA).

NGS

After 8 selection rounds, the C2 starting library and samples from round 5 and 8 were prepared for next generation sequencing (NGS) analysis on NextSeq500 with High Output v2 chemistry (Illumina) following the protocol from Tolle et al. [16]. Starting library and selection rounds were amplified with index primers synthetized by Ella Biotech GmbH (Munich, Germany). A PCR reaction mix was prepared for each pool and for an additional No Template Control (NTC), using a proofreading Pwo-Polymerase (Genaxxon bioscience, Germany). 1 µM of forward and 1 µM of reverse primer with the same index sequences were added. Adding 1 µL of dsDNA or 1 µL of double distilled H2O as NTC, the dsDNA was amplified by PCR (3 min 95 °C, 60 s 95 °C, 60 s 56 °C, 60 s 72 °C, 3 min 72 °C, hold 10 °C). PCR product was purified with Nucleospin® Gel and PCR clean-up columns (Macherey & Nagel, Germany) according to the manufacturer’s manual. 0.125 µg of purified DNA for each sample was used for ligation of the adapter needed for immobilization and processing of the sample on the NextSeq500 using TruSeq DNA PCR-Free Sample Preparation Kit LT (Illumina, USA). Ligation was done according to instructions of manufacturer applying only three procedures: “End Repair”, “Adenylation” and “Adapter Ligation”. Afterwards, the ligated products were separated on a 2% agarose gel, cutting out the longest sequence (∼210 bp) with ligated adapters on both sides. The DNA was purified from the gel using Nucleospin® Gel and PCR clean-up columns, following manufacturer instructions. DNA was resuspended in resuspension buffer (TruSeq DNA PCR-Free Sample Preparation Kit LT) and was used for Ilumnia sequencing. NGS analysis was done with the in house developed software.