key: cord-1051621-exoxy64h
authors: Barthels, Fabian; Hammerschmidt, Stefan J.; Fischer, Tim R.; Zimmer, Collin; Kallert, Elisabeth; Helm, Mark; Kersten, Christian; Schirmeister, Tanja
title: A low-cost 3D-printable differential scanning fluorometer for protein and RNA melting experiments
date: 2022-01-07
journal: HardwareX
DOI: 10.1016/j.ohx.2022.e00256
sha: cbf935027a8ac89bf5ba27682da4076a30cee5f1
doc_id: 1051621
cord_uid: exoxy64h

Differential scanning fluorimetry (DSF) is a widely used biophysical technique with applications to drug discovery and protein biochemistry. DSF experiments are commonly performed in commercial real-time polymerase chain reaction (qPCR) thermal cyclers or nanoDSF instruments. Here, we report the construction, validation, and example applications of an open-source DSF system for 176 €, which, in addition to protein-DSF experiments, also proved to be a versatile biophysical instrument for less conventional RNA-DSF experiments. Using 3D-printed parts made of polyoxymethylene, we were able to fabricate a thermostable machine chassis for protein-melting experiments. The combination of blue high-power LEDs as the light source and stage light foil as filter components was proven to be a reliable and affordable alternative to conventional optics equipment for the detection of SYPRO Orange or Sybr Gold fluorescence. The ESP32 microcontroller is the core piece of this openDSF instrument, while the in-built I(2)S interface was found to be a powerful analog-to-digital converter for fast acquisition of fluorescence and temperature data. Airflow heating and inline temperature control by thermistors enabled high-accuracy temperature management in PCR tubes (±0.1 °C) allowing us to perform high-resolution thermal shift assays (TSA) from exemplary biological applications.

We designed an open-source differential scanning fluorometer as a low-cost alternative to existing qPCR thermal cyclers and commercial nanoDSF devices for biophysical characterization of protein-ligand complexes (Fig. 1 ). To support lowbudget academic research campaigns, only inexpensive and easily available components (e.g., 3D-printed pieces, basic electronics) were used, so that the costs (176 €) are several magnitudes lower than those of a commercial counterpart ($20,000 €) [17] . The fundamental principle of the system is that four DSF reaction mixtures, prepared in PCR tubes, are flowed with heated air, while both the temperature and the SYPRO Orange fluorescence of each PCR tube are measured and recorded. To achieve this functionality, the design of the device combines five modules to form the whole open-source differential scanning fluorometer: 1) the 3D-printed chassis, 2) the ESP32 microcontroller, 3) the air heating system, 4) the temperature probes, and 5) the fluorescence measurement path (Fig. 2B ).

Structural parts forming the chassis of the device were fabricated by fused deposition modeling with an Anycubic 4Max 3D-printer either from polyethylene terephthalate (PETG, diameter: 1.75 mm) or from polyoxymethylene (POM, diameter: 1.75 mm) filament. Functional tasks of the chassis are the assembly of the measuring, lighting, and heating modules as well as the directional flow control of the air heating system. DSF experiments are performed in the temperature range of 20-95°C, which sets special requirements for the temperature stability of the 3D-printed parts. Hence, POM filament was used for parts with increased temperature demands (POM is dimensionally stable up to at least 130°C [18] ). To meet the known constructional limitations of POM 3D-printing, sharp corners and edges were avoided in the design of the POM parts, because these are particularly prone to warping [19] . The openDSF chassis consisted of 14 different 3D-printed parts ( Fig. 2A) . POM parts were designed double-walled and printed with an infill of 20% so that the high air content inside these parts makes them good thermal insulators. The low thermal losses allow for reproducible heating ramps and eliminate hot zones on the device surface so that it can be safely touched by the user at any time.

The Espressif ESP32 wrover microcontroller (dual-core CPU, 8 MB PSRAM, 240 MHz clock) was chosen as the central control unit of the fluorometer. The ESP32 is a low-cost and relatively recent microcontroller family (released in 2016) [20] . The digital outputs of the chip were used to control the excitation LEDs, the air-supplying fan, and the heating wire. The peripheral I 2 S interface of the chip, which has been designed for transmitting audio data, has proven suitable for recording thermistor and photodiode analog input signals in the microsecond range (Fig. 2B) . In this regard, two onboard multi-channel successive-approximation analog-to-digital converters (SAR-ADCs) were polled with a sampling rate of 60 kHz and the temperature resp. fluorescence values were recorded via direct memory access (DMA) protocol. Online data processing was handled by one of the two CPU cores, while the other core was tasked with general control and communication via the PC's serial interface. The firmware for the ESP32 was developed on the Espressif ESP-IDF 4.2 platform, while the serial interface was implemented with Python, Matplotlib, NumPy, SciPy, and Tkinter. A serial/USB driver must be installed on the connected PC.

Temperature control of the sample PCR tubes was performed by airflow heating, in which the air was tempered with a heating wire and conveyed by a radial fan (Fig. 3A) . The heating power (heating wire current) was controlled by a 13-bit PWM MOSFET circuit, while the air volume (fan speed) was controlled by an 11-bit circuit. A Sunon Blower (MF50151V2- B00U-A99) was used as a radial fan, which was designed to convey approx. 5 m 3 /h of air along an SS430 awg30 heating wire (stainless steel 430, l = 70 cm, d = 0.25 mm, R = 12.2 X/m, X6Cr17). The power consumption of the heating module was designed to be max. 150 W to be able to run both slow and fast (0.1-9°C/min) heating ramps for up to 4 PCR tube samples. A homogeneous temperature cross-section was ensured by turbulent airflow within the device's mixing cell. The dimensions of the air-flowed round parts of the chassis were designed based on a calculated Reynolds number (Re $ 10,000). Additionally, to increase the velocity of the airflow and thus the heat transfer in the proximity of the sample vessels, the sample holders were shaped as Laval nozzles. The resulting pressure loss promotes uniform distribution of the airflow to the 4 individual sample holders.

The temperature of each sample can be determined separately using negative temperature coefficient (NTC) thermistors (NTC3950, 100 kX) within the airspace of the individual PCR tube. The temperature probes were dimensioned in a manner that they can be inserted into the PCR tube and thereby seal the upper opening. A thermistor circuit was designed in such a way that a voltage divider (V ref = 5.02 V) enabled the recording of relevant measurement temperatures (20-95°C) within the ADC's 12-bit resolution. The measured temperature-dependent voltages were recorded every second by the ESP32 chip and stabilized by oversampling (SAR-ADC at 60 kHz sampling rate). The conversion of voltages to sample temperatures was performed using the modified Steinhart-Hart equation ( 1 T ¼ a 0 þ a 1 ln R ð Þ þ a 2 ln 3 ðRÞ) [21] . The apparent resistance was calculated from the measured voltages and the dimensions of our design: R thermistor ¼ 12kXð 5;02V 0:0008ÁUþ0:0624 À 1Þ . The thermistors were calibrated by determining the coefficients (a i ) using a precision contact thermometer (Pt100). With this setup, we were able to achieve linear heating ramps over the entire temperature range, with an absolute temperature inaccuracy of ± 0.1°C (Fig. 3B ).

The conventional DSF dye SYPRO Orange is a merocyanine-type fluorescent dye with absorption and emission maxima of k ex = 490 nm and k em = 624 nm (Fig. 4A ). Commercial qPCR instruments often use costly lamps, optics, or filter equipment, which we have circumvented by using cheap high-power OSRAM Oslon SSL 80 royal-blue LEDs as the light source (on 20 mm stars, LD CQ7P). The fluorescent dye in the samples was excited with a 50 ms pulse of the high-power blue LED (k = 450 nm, W max = 1400 mW, / E = 630 mW). Longer pulse widths resulted in significant photobleaching of the dye during long-term experiments, but since the fluorescence signal is recorded at a sampling rate of 60 kHz, a 50 ms pulse width was found to be sufficient to achieve low-noise recordings. Since the output fluorescence depends on the protein, the buffer, and the sample volume, the irradiation intensity can be modulated with the 11-bit PWM controller of a constant current source (LED BUCK V2, 1000 mA, 42 V, PWM < 5 kHz). Excitation of the sample solution was performed through the lateral outer wall of the PCR tube, with the emitted fluorescence captured at a 90°angle by OSRAM SFH 203 photodiodes (Fig. 4B ). The SFH 203 photodiode was selected because of its exceptionally fast switching operational performance (rise/fall time: 5 ns). However, this photodiode exhibits sensitivity throughout the entire visible spectrum, so that the scattered and reflected excitation light must be eliminated with a filter. We have found that filter foils for stage spotlights are an excellent alternative to expensive optical filters. In this regard, the receiver photodiode was shielded with a piece of the LEE 158 Deep Orange filter foil, which efficiently prevented the passage of light with a wavelength of < 500 nm. The current of the receiver photodiode was evaluated with a transimpedance amplifier (Texas Instruments OPA380, R F = 12 MX) and the resulting output voltage was measured by the ESP32 SAR-ADC (sampling rate of 60 kHz, bit depth of 12-bit).

During a DSF run, both fluorescence and temperature data are acquired for each of the four holders individually (ADC serial inputs: 8 channels = 4 each for photodiodes and thermistors). In this regard, two values per data point, i. e. temperature and fluorescence, are recorded for regular intervals. The interval between two measurement points has been established with 1000 ms by default but can also be modified (def TEMP_SAMPLE_PERIOD). At the beginning of each interval, a sampling phase is performed in which the PCR tube is irradiated, and fluorescence resp. temperature values are recorded. The default duration for the sampling phase is determined by the LED pulse width (int led_on; default: 50 ms). The sampling frequency is 60 kHz on 8 channels (4 temperature, 4 fluorescence), hence, for each data point 375 raw values per channel (¼ 60kHz 4þ4channels Á 50ms 1000ms ) are recorded. Subsequently, these raw values are combined by arithmetic averaging and the mean fluorescence values are written to a log file for online plot and CSV export at specified temperatures (0.1-1°C/datapoint). For post-acquisition processing, a smoothing function according to the Savitzky-Golay filtering algorithm was implemented (scipy.signal.savgol_filter). The signal acquisition and processing pipeline of the openDSF system is shown in SI Fig. 1 .

The open-source differential scanning fluorometer (openDSF) allows fast, accurate, and reliable protein and RNA-DSF measurements. The combination of high-power LEDs as the light source and stage light foil as filter components is a cost-effective alternative to laser lamps and optics equipment for fluorometric measuring cells. Airflow heating and temperature control by thermistors allow high accuracy temperature management in PCR tubes (±0.1°C). The ESP32 microcontroller with the I 2 S interface is a powerful system for fast acquisition of analog measurement data (sampling rate 60 kHz). A 3D-printable polyoxymethylene (POM) chassis allows biological experiments at up to 130°C.

The repository containing the design data for reproduction and modification of an openDSF instrument can be found at https://doi.org/10.17632/73rt8s7pwd. AirDiffuser: The air diffuser (POM) divides the heated airflow among the four individual measuring cells and leads to turbulent air mixing and thus to a homogeneous temperature profile.

DiodeShield: The diode shield (PETG) is designed to shield the photodiodes from external light or daylight to reduce the noise of the recorded fluorescence signal.

FanAdapter: The fan adapter (PETG) converts the rectangular outlet of the Sunon radial fan to a circular fitting of the heating tube.

FluorescenceCell: The fluorescence cell (PETG) creates a framework into which the sample holders are mounted.

HeatingTube: The heating tube (POM) incorporates the coiled heating wire.

Housing: The housing of the circuit boards protects the electronic components during use.

LEDHolder: High-power LEDs are glued onto the LED holders and serve to stabilize them. openDSF: Project folder (openDSF.zip) for insertion into an ESP_IDF development environment. Python GUI (openDSF.py) to control the device from the user's PC.

PhotoDiodePCB: Material for the fabrication of printed circuit boards of the photodiode modules.

SampleCover: A light and temperature shielding cover (PETG + POM) is placed over the measuring cell, through which the temperature sensors are inserted into the PCR tubes. The heated airflow of the instrument exits through this cover, which is why the contact points to the hot air were protected with thermally robust POM inserts.

SampleHolder: The sample holders (POM) are shaped as Laval nozzles to ensure optimal airflow and heat transfer. Due to the pressure loss of the nozzle, homogeneous air distribution is achieved. The sample holders provide holes for the LEDs and photodiodes.

TemperatureSensor: The temperature sensors (PETG) hold the individual thermistors and shield the sensitive electrical contacts from moisture and physical contact.

WiringDiagram: Circuit diagram overview of the electronic components of the openDSF system created with KiCad Eeschema.

3D-printing profiles: List of 3D-printer settings (.ini) for the different filaments used (PETG and POM). Build instructions

Never heat or turn on the LED lights while the openDSF instrument is disassembled. Although the device uses low voltages (36 V), hot parts are inside the apparatus (heating wire). When assembled correctly, the high-power LEDs are directed into the interior of the apparatus, so they emit very little stray light to the environment. However, looking directly into the disassembled LEDs can irritate the user's eyes. The SYPRO Orange resp. Sybr Gold DSF dye does not have a hazardous material classification, however, the biological samples and chemicals used in individual DSF experiments may harbor specific hazards. We recommend that the openDSF instrument, like all heating laboratory equipment, should not be left running unattended for several hours as they pose a potential fire risk in the event of a malfunction.

We fabricated the 3D-printed parts with an Anycubic 4Max printer. Utilizing the open-source software FreeCAD 0.18 and Ultimaker Cura 15.04.6 39, we designed, meshed, and sliced individual components. We performed printing through a 0.4 mm nozzle with 1.75 ± 0.02 mm PETG filament at 225°C. The layer thickness was set to 0.2 mm. POM was printed at 240-250°C and a bed temperature of 70°C. The infill percentage and infill style were 20% resp. the zig-zag pattern for both filaments. The complete printer settings were added as Cura profiles (.ini) to the repository of design files. An overview of all 3D-printed components can be seen in Fig. 2 . For better adhesion to the printing table, POM was printed on a three-layer bed of PETG filament. A layer printing time of 6-8 s/layer was found to be most effective against warping effects for the POM filament.

Winding of the heating tube:

In the POM heating tube, a coil of the heating wire (8 mm diameter with 28 windings corresponding to a total length of approx. 0.7 m; pitch approx. 1.8 mm) was inserted (Fig. 3A) . The ends of the coil were screwed onto cut-to-size drilled board parts (24 Â 5 mm). To fix the coil vertically, the drilled board parts were clamped in the inlet and outlet of the heating tube, respectively. The power connections were soldered on the drilled boards and the connections were led out through 3 mm drilled holes in the heating tube.

Before the 3D-printed parts, the heating tube, and the fan were assembled by their plug-in joints, the inner surfaces of the colorless POM parts (diffuser, sample holder, and LED holder) were coated with a black polyurethane varnish layer to absorb stray light. Alternatively, the corresponding parts might be printed with a non-transparent material, e.g., black filament. Previously, we also found that the mechanical durability of varnished 3D-printed parts is increased compared to the raw form [22] . Subsequently, a rolled piece of orange filter foil was inserted into the four individual sample holders and the foil was fixed with superglue ( Fig. 5A) . Caution: The filter foil is attacked by the contained solvents.

The individual parts of the openDSF instrument were assembled according to the exploded view drawings ( Fig. 2A and  4B ). Note: Make sure that the joints are well sealed! Any printing imperfections must be corrected by filing. The critical plug connections (fan -fan connector -heating tube -diffuser) can optionally be sealed with Teflon insulating tape. To facilitate rebuilding, we have added a 3D model of the exploded view drawing from Fig. 2A to the repository of design files (https://data.mendeley.com/public-files/datasets/73rt8s7pwd/files/90f0a996-645d-4283-8ca8-ae1a65b3404b/file_downloaded), so that a rebuilder can examine the topology of the design while zooming in or moving parts around.

For assembly of the fluorescence cell, two LED modules and two photodiode modules were fabricated, so that a sample holder is equipped with one LED and one photodiode each (Fig. 4B) . For a single LED module, two high-power LEDs were soldered together with proper polarity and provided with connecting cables so that they could be glued onto their LED holder with hot glue and inserted into the cut-outs at the fluorescence cell. To assemble the photodiode module, two photodiodes were mounted (20 mm apart from each other) on a PCB board according to the PCB layout specifications (https://data.mendeley.com/public-files/datasets/73rt8s7pwd/files/81dcaa9e-9a6d-44c0-a272-a710958638ab/file_downloaded).

We manufactured the PCBs ourselves using the toner transfer method [23] . The electronic components were soldered onto the circuit board using a standard soldering iron. For the operational amplifier, assembly instructions from the manufacturer's datasheet are to be followed. The photodiodes were inserted into the holes provided in the fluorescence cell and later covered by the two PETG diode shields (Fig. 4B) . Wiring of electronic components:

The thermistors were inserted into the 3D-printed temperature sensors and fixed so that the glass sphere of the thermistor just peeks out of the bottom hole (Fig. 6B) . The MOSFET circuits (fan and heater) were assembled on drilled boards. A flyback diode was provided for the fan. The resistance values from the PWM supply were 15 X and 10 kX to the ground. Subsequently, the ESP32 development board, the constant current power supply, and the MOSFET circuits were placed in the housing (Fig. 5B) . Finally, the individual electronic components were cabled according to the circuit diagram (https://data.mendeley.com/public-files/datasets/73rt8s7pwd/files/2dbbf304-9985-4c7c-988e-8e972766ae3e/file_downloaded). We noticed that using a USB cable without supplying voltage between the ESP32 and the computer leads to more stable ADC characteristics. The possible interference between simultaneous supply via the USB voltage common collector (VCC) and the 5 V pin of the ESP32 board has already been described for V ext < 5.2 V [24] . Therefore, we prepared a dataonly USB cable from a standard micro-USB cable for data transfer to the computer according to literature instructions [25] . The archive ''openDSF.zip" (https://data.mendeley.com/public-files/datasets/73rt8s7pwd/files/ba2f88ef-547c-4ca7-8ef7-ed597cbf53b8/file_downloaded) has been unpacked. The unpacked openDSF folder was copied into the root directory of the development environment. Before the firmware can be compiled, the development environment was slightly modified. The file ''. . .

Evaluation of fluorescence-based thermal shift assays for hit identification in drug discovery

High-density miniaturized thermal shift assays as a general strategy for drug discovery

Three Essential Resources to Improve Differential Scanning Fluorimetry (DSF) Experiments, preprint (BioRxiv)

Differential Scanning Fluorimetry as Part of a Biophysical Screening Cascade

Theory and applications of differential scanning fluorimetry in early-stage drug discovery

Isothermal analysis of thermofluor data can readily provide quantitative binding affinities

The use of differential scanning fluorimetry to detect ligand interactions that promote protein stability

SimpleDSFviewer: A tool to analyze and view differential scanning fluorimetry data for characterizing protein thermal stability and interactions

Optimal filter combinations for photographing SYPRO orange or SYPRO red dye-stained gels

Low cost, microcontroller based heating device for multi-wavelength differential scanning fluorimetry

An affordable and portable thermocycler for real-time PCR made of 3D-printed parts and off-theshelf electronics

3D printing and milling a real-time PCR device for infectious disease diagnostics

Quantitative Real-time PCR Analysis

Melting studies of RNA unfolding and RNA-Ligand interactions

Synthetic RNA modulators in drug discovery

Differential scanning fluorimetry for monitoring RNA stability

International Plastics Handbook

Atmospheric pressure air plasma treatment to improve the 3D printing of polyoxymethylene

Using the ESP32 Microcontroller for Data Processing

User-friendly calibration tool for temperature measurements of PCR devices with NTC thermistors

FINDUS: An open-source 3D printable liquid-handling workstation for laboratory automation in life sciences

Heatless (cold) Toner Transfer for PCB Making, Instructables

Analysis and Optimization of DC Supply Range for the ESP32 Development Board, preprint (TechRxiv)

Extraction of thermodynamic parameters of protein unfolding using parallelized differential scanning fluorimetry

Determination of protein-ligand interactions using differential scanning fluorimetry

Meltdown: A tool to help in the interpretation of thermal melt curves acquired by differential scanning fluorimetry

Modulated scanning fluorimetry can quickly assess thermal protein unfolding reversibility in microvolume samples

Thermal shift assay to probe melting of thrombin, fibrinogen, fibrin monomer, and fibrin: Gly-Pro-Arg-Pro induces a fibrin monomer-like state in fibrinogen

GC-376, and calpain inhibitors II, XII inhibit SARS-CoV-2 viral replication by targeting the viral main protease

Asymmetric disulfanylbenzamides as irreversible and selective inhibitors of Staphylococcus aureus sortase A

Optimization strategy of single-digit nanomolar cross-class inhibitors of mammalian and protozoa cysteine proteases

Calpain activation by cooperative Ca2+ binding at two non-EF-hand sites

Cis autocatalytic cleavage of glycine-linked Zika virus NS2B-NS3 protease constructs

Boenitz-Dulat, I. Antes, Ca2+ binding induced sequential allosteric activation of sortase A: An example for ion-triggered conformational selection

Conversion of an amide to a high-energy thioester by Staphylococcus aureus sortase A is powered by variable binding affinity for calcium

A Ca2+ switch aligns the active site of calpain

Crystal structures of Calpain-E64 and -leupeptin inhibitor complexes reveal mobile loops gating the active site

BANDIT: B'-factor analysis for drug design and structural biology

Solution conformations of a linked construct of the Zika virus NS2B-NS3 protease

Structure-based macrocyclization of substrate analogue NS2B-NS3 protease inhibitors of Zika, West Nile and Dengue viruses

Structural dynamics of zika virus NS2B-NS3 protease binding to dipeptide inhibitors

Impacts of fluorescent base analogue substitution on the folding of a riboswitch

Post-SELEX chemical optimization of a trypanosome-specific RNA aptamer

During his doctorate, he discovered his interest in open-source hard-and software for drug discovery applications. Previously, he has developed and published an open-source liquid handling workstation (FINDUS) and a crystallographic Bfactor analysis toolkit (BANDIT)

Research in the laboratories of Mark Helm and Tanja Schirmeister was funded by DFG grants (TRR 319 ''RMaP" projects A01, C01, and A05). Cruzain was a gift from the group of Dr. Avninder Bhambra (University of Leicester, UK). We thank Prof. Torsten Steinmetzer (University of Marburg, Germany) for providing the cyclic peptide ligand (1-((8R,15S,18S)-15,18-bis(4aminobutyl)-4,7,14,17,20-pentaoxo-3,6,13,16,19-pentaaza-1(1,3)-benzenacyclohenicosaphane-8-yl)guanidine).

Supplementary data to this article can be found online at https://doi.org/10.1016/j.ohx.2022.e00256.