2LabsToGo—Recipe for Building Your Own Chromatography Equipment Including Biological Assay and Effect Detection

Abstract

A complete recipe for building your own chromatography equipment from readily available materials is introduced. It combines sample separation (chemistry laboratory) with biological effect detection (biology laboratory). This hyphenation of two disciplines is necessary for prioritizing important compounds in complex samples. Among the thousands of compounds therein, it is often not clear which compounds are the important ones. On the same separation surface, additional detection of biological effects enables and guides substance prioritization. The newly developed open-source 2LabsToGo system for chemical and biological analysis is completely solvent-resistant and, due to miniaturization, environmentally friendly regarding the consumption of materials. It produces comparable results but is 10 times more compact (26 cm × 31 cm × 34 cm), 10 times lighter (6.8 kg), and 55 times less expensive (€ 1717) than current sophisticated commercial devices. As a proof of concept of the first 2LabsToGo system, the quality of different water samples was analyzed since clean water is becoming increasingly rare. In water, most of the thousands of substance signals or features can neither be identified nor classified toxicologically. However, methods that exploit this hyphenated strategy provide answers to such essential safety issues. Drinking or tap water did not show bioactive or toxic compounds, which was expected, whereas biogas or landfill water samples did. The hyphenated 2LabsToGo strategy is affordable and extremely useful for all laboratories with limited equipment but pressing challenges. It is ready to be used in various analytical tasks and applications.

Additive manufacturing has become an integral part of open-source, low-cost, and self-manufactured operational steps.¹⁻³ However, it has hardly been used for building complete chromatographic systems.⁴ Even beyond that, 2LabsToGo systems have not yet been demonstrated. The combination of chromatographic separation with imaging of individual biological effects in complex samples is a powerful emerging field.⁵ Since modern analysis lacks prioritization of the crucial substances from the thousands of compounds in a complex sample, it can become a game-changer for solving pressing analytical challenges. For example, up to 20 000 different signals are detected in water using sophisticated multidimensional analytical systems. At present, no one can answer what most of these substance signals mean because they are mostly unidentified signals.⁶ In addition, there are no data available on the toxicological potential of each detected signal, and the in vitro assays (which only give a sum parameter result) have proven to be misleading regarding opposite effect results in complex samples.^7,8 Since clean water is becoming increasingly rare (https://unstats.un.org/sdgs/report/2021), the challenge is evident. Plant-based food and other environmental samples are expected to be even more complex than water.

Structure elucidation and toxicological studies cannot be performed for thousands of unknown compounds in a sample that also exhibits a natural variance in its composition. Hence, combining two disciplines offers pragmatic solutions to difficult analytical tasks. In particular, the fusion of chromatographic separation and biological assay detection is helpful for the decision-making process on a complex sample. However, column chromatography systems combined in flow with bioassays are not compatible with the long bioassay incubation times required. Although planar chromatography can ideally be combined with bioassays, the instrumentation is very expensive. This hinders the rapid spreading of the application of effect-directed analysis, which, however, enables prioritizing the crucial substances in a complex sample.

The progress in novel technologies inspired us to build all-in-one miniaturized equipment driven by the principles of Simplicity, Miniaturization, Do-it-yourself, Additive Manufacturing, Open Source Technologies, Lean Laboratory, and Sustainability (Green Chemistry). Our first approaches to developing a miniaturized chromatography system led to the demounting of low-cost inkjet printers,^9,10 then to the integration of open-source technologies,^11,12 still based on the inkjet cartridge. However, the previous OCLab systems were severely limited in solvent compatibility, applicability, and performance (they did not cover all desired steps). Although sample images can be captured in a 3D-printed miniature cabinet illuminated by light-emitting diodes (LEDs) via an inexpensive Raspberry Pi mini-camera^11,12 or smartphone camera,¹³ the important bioluminescence detection of planar bioautograms to capture the biological effect has not yet been shown on a miniaturized scale. Although effect images can be evaluated digitally using free-of-charge open-source software for videodensitometry,¹⁴ multivariate data analysis,¹⁵ and artificial neural networks,^16,17 further research and advancement are required for intuitive or even unsupervised operation of the image evaluation.

In this study, we introduce, for the first time, a completely solvent-resistant 2LabsToGo system and the recipe for building it from readily available materials on an open-source basis. It was and can be self-built by analytical chemists, which in turn provides an understanding of its operation and repair. The first proof of concept is exemplarily given for the analysis of a food dye mixture and different water samples. However, many applications and workflows, so far handled with commercial instruments, can be replaced by the new 2LabsToGo system, such as the analysis of food composition,¹⁸⁻²⁰ contaminants,^21,22 and residues,²³ as well as the bioprofiling of plants,²⁴⁻²⁶ foods,^7,8,27,28 and cosmetics.²⁹

Experimental Section

Chemicals and Materials

Bidistilled water was produced by Heraeus Destamat Bi-18E, Thermo Fisher Scientific, Schwerte, Germany. Organic solvents (chromatography grade) were delivered by VWR, Bruchsal, Germany. Crystal Ponceau 6R (E 126), Patent Blue V (E 131), Brilliant Scarlet 4R (E 124), Erythrosine (E 127), Fast Yellow AB (E 105), Tartrazine (E 102), methylparaben, and caffeine were purchased from Sigma-Aldrich, Taufkirchen, Germany. HPTLC plates silica gel 60 NH₂ F₂₅₄, silica gel 60 F₂₅₄, and MS-grade silica gel 60 F₂₅₄, all 10 cm × 10 cm, were from Merck, Darmstadt, Germany. The bioluminescent marine Gram-negative Aliivibrio fischeri bacteria were obtained from DSM-7151, German Collection of Microorganisms and Cell Cultures, Berlin, Germany. The medium is listed elsewhere.³⁰ Nine different types of water samples were collected. Luminous fluorescent tape (10 m × 15 mm, TANCUDER, amazon.com) was used for studying the camera for long-exposure images.

Samples

The water-soluble food dyes were dissolved in water as follows: E 126 (26 ng/μL), E 131 (28 ng/μL), E 124 (21 ng/μL), E 127 (20 ng/μL), E 105 (112 ng/μL), E 102 (42 ng/μL); aliquots of 200, 200, 150, 70, 500, and 200 μL, respectively, were mixed to obtain the dye mixture solution. Water samples were centrifuged at 3000g for 15 min or filtered (0.22 μm filter) and stored at −20 °C until use. As positive controls for the bioassay, caffeine and methylparaben solutions (1 μg/μL each in methanol) were prepared.

Hardware

For self-mounting, assembly instruction is available (Instructions S1 and S2). The x-profiles of the frame (Motedis AGB, Endorf, Germany) were made of aluminum. All 3D prints of system parts were designed in FreeCAD (freecadweb.org) or OpenSCAD (openscad.org) and are freely available for download at https://github.com/OfficeChromatography/OCLab3. Polylactic acid was used to create the 3D-printed parts by a Prusa i3 MK3 3D printer (Prusa Research, Praha, Czech Republic). As an exception, the plate holder was 3D-printed using the more heat-resistant (max. 140 °C) polycarbonate blend carbon fiber, and the miniature cabinet was printed with matt polylactic acid (both www.filamentworld.de). Plate images were captured by the rather new Raspberry Pi 12 MP HQ camera with a 6 mm wide-angle lens mounted onto a newly developed miniature cabinet. Two NeoPixel sticks (www.adafruit.com/product/2869) were used, each with eight RGBW LEDs addressable by the driver chip inside the LEDs for white light illumination. For UV light illumination, four UVA 365 nm (EDC365V-1100-Z, Marubeni Europe, Düsseldorf, Germany) and four UVC 265 nm LEDs (QBHP684E-UV265N, Digi-Key Electronics, Thief River Falls, MN) were installed, soldered onto respective printed circuit boards (PCB) controlling the currents. A separate PCB was constructed for the LED power supply (Instruction S1).

Firmware

Created files are available at https://github.com/OfficeChromatography/OCLab3. The apparatus was controlled by an Arduino Mega 2560 microcontroller board (https://arduino.cc) associated with a RAMPS 1.4 shield (http://reprap.org/wiki/RAMPS_1.4). Multiple new Gcodes were added to the Marlin firmware (http://marlinfw.org) to control the machine, creating the new OC-Control firmware.

Software

The open-source software OC-Manager3, newly developed in Python, is freely available at https://github.com/OfficeChromatography/OC-Manager3. Its operation is described in a user manual in the Supporting Information (Instruction S3). The OC-Manager3 manages the generated Gcodes and utilizes the Django framework, Bootstrap, and jQuery to create a user interface. The Printcore module was used (http://www.pronterface.com) to communicate with the Arduino. The software was hosted on a Raspberry Pi 4 single-board computer (https://raspberrypi.org) connected to a local network.

Food Dye Analysis

The food dye mixture solution consisting of six different food dyes as well as the respective mobile phase were ejected on an HPTLC plate silica gel 60 F₂₅₄. The syringe was filled and rinsed with sample solution or mobile phase and placed into the syringe pump. The plate holder was heated up to 70 °C, and the heat was maintained during application to support the evaporation of the solvent. For the sample application, 20 bands of 1.3 mm length were applied with a lateral distance of 2 mm, and distances to the lower and left plate edges of 12 mm and 5 mm, respectively. A distance of 0.6 mm between drop applications was used to generate the band. The ejection pressure was 5 psi, and the valve opening frequency was 1400 Hz. The volume was set to 0.375 μL for bands 1–5 and increased by 0.075 μL increments up to 1.5 μL for bands 6–20. For development, the plate holder was closed via the glass cover and glass stripe. The mobile phase mixture consisting of ethyl acetate–methanol–water–acetic acid, 65:23:11:1, V/V/V/V,³¹ was ejected by nozzle movement (speed 25 mm/s) as a forward-backward-line along the 10 cm plate using an ejection pressure of 30 psi. The distances to lower, left, and right plate edges were set to 2, −10, and −10 mm, respectively. Twenty passes of 75 μL/s with a waiting time of 20 s were performed. The migration distance was 35 mm, and the plate development took 4.5 min. The plate holder was again heated up to 100 °C for 1 min to let evaporate the mobile phase solvents. The dried chromatogram was documented at white light LED illumination using the built-in Raspberry Pi 12 MP HQ camera with a 6 mm wide-angle lens. The images were digitally preprocessed, selecting negative peak inversion, smoothing, baseline correction, and warping. Videodensitograms were automatically generated and evaluated via the open-source software quanTLC (http://shinyapps.ernaehrung.uni-giessen.de/quanTLC).¹⁴ Automatic peak integration was performed (minimum number of in-/decreasing steps before/after the peak was set to 1).

Bacterial Cultivation

The bioassay medium³⁰ was inoculated with nonpathogenic marine A. fischeri bacteria and incubated at room temperature in a baffled vessel on an orbital shaker at about 30 rpm (Behr, Düsseldorf, Germany) overnight (ca. 18 h). The bacteria are rather robust concerning the mode of cultivation. The culture′s readiness can easily be checked in a dark room by shaking the vessel. When brilliant blue-green bioluminescence is observed, it is ready to be dosed on the chromatogram. The culture could be used for a week when stored cooled overnight or/and adding a portion of fresh medium.

Water Analysis via OCLab3

The syringe was filled and rinsed with the respective solution and placed into the syringe pump. The plate holder was heated up to 90 °C, and the heating was maintained during application to support the evaporation of the water. Nine different water samples, as well as focusing solvent and mobile phase for its separation, were ejected on an HPTLC plate silica gel 60 F₂₅₄ MS-grade. For 50 μL water sample applications, an ejection pressure of 40 psi and a valve opening frequency of 1200 Hz were set. A top/side distance of 0.8 mm between drop applications was used to form the area. Nine areas of 5 mm × 15 mm were applied with a lateral distance of 5 mm, and distances to the lower and left plate edges of 12 mm and 7.5 mm, respectively. For focusing (optionally two times), the mixture of acetone and water 1:1, V/V, was ejected at a speed of 25 mm/s by nozzle movement in both directions up to a migration distance of 27 mm using an ejection pressure of 40 psi. The distances to lower, left, and right plate edges were set to 2, −10, and −10 mm, respectively (same for development). Six passes of 12 μL/s with a dynamic waiting time were performed. For development, the nozzle speed movement was set to 35 mm/s and the ejection pressure to 20 psi. The mobile phase mixture consisting of ethyl acetate–toluene–methanol–water 65:20:11:4, V/V/V/V, was ejected up to a migration distance of 70 mm. Forty passes of 12 μL/s with a dynamic waiting time were performed. The plate holder was again heated up to 100 °C for 3 min to let evaporate the mobile phase solvents. For documentation of the UV/Vis/FLD chromatograms, the plate was illuminated at white light, 265 nm, and 365 nm LEDs, and the image was captured via the Raspberry Pi 12 MP HQ camera with a 6 mm wide-angle lens. As a positive control for the bioassay, a methylparaben solution (0.5–4 μg/band) or a caffeine solution (1–5 μg/band, in the upper plate edge of the water analysis plate) were applied. Bioluminescent plate images were captured as described in Instruction-S2.

Comparison to Commercial Instrumentation

Equivalent commercial instruments were considered for the comparison concerning investment costs, laboratory bench space, and weight (Table S1), i.e., the Automated TLC Sampler (ATS 4) for sample application, Horizontal Developing Chamber (HDC, 10 cm × 10 cm) for development, TLC Plate Heater for solvent evaporation, TLC Visualizer 2 for documentation, Derivatizer for bioassay application, BioLuminizer for bioluminescence detection, and visionCATS Ultimate software for digital evaluation and operation (all CAMAG, Muttenz, Switzerland).

Results and Discussion

Features and Construction

The developed portable OCLab3 being the first open-source 2LabsToGo system is considered a milestone (Figures 1 and S1). The all-in-one small-scale planar separation system is now compatible with Citizen Science and can be used in a wide range of applications. For the first time, it provides full solvent compatibility and is enhanced and enriched in important additional functionalities, such as chemical derivatization, bioassay application, mini-incubator, and bioluminescence detection (BD). It combines sample separation (chemistry laboratory) with biological effect detection (biology or forensic or toxicological laboratory), making it the first 2LabsToGo system able to prioritize important compounds in complex samples. The 2LabsToGo system is controlled via an intuitive graphical user interface and, for the first time, programmed in the language Python. This makes it connectable to unsupervised cloud-based image analysis. All system parts can readily be bought from manufacturers (Table S2) or 3D-printed (open-source files available at https://github.com/OfficeChromatography/OCLab3) and then self-mounted via step-by-step assembly instructions (Instructions S1 and S2; depending on skills, it may take 2–3 days for analytical chemists to mount it).

Figure 1.

Developed solvent-resistant, open-source 2LabsToGo system, consisting of a syringe pump (1), a valve-nozzle system (2), an electronics compartment (3), a miniature cabinet with a camera and LEDs (4), a multifunctional heatable plate holder (5), and motors for positioning (6).

The electronics (Instruction S1) consists of a Raspberry Pi 4 Model B with 8 GB RAM to host the newly coded open-source software OC-Manager3 (Instruction S3, free download at https://github.com/OfficeChromatography/OC-Manager3). An Arduino Mega 2560 was connected via a USB cable to run an instance of a modified version of the Marlin firmware. The Arduino board controls the electromechanical devices. A Ramps shield was plugged on top of it to provide the power to command the actuators (motors, heating bed, fan, sensors, and LEDs). The three boards were placed on top of the system (Figure 1, part 3). Furthermore, three endstops were connected to it to calibrate the motor movements. The operation of system parts (of a former prototype) is explained in a 10 min video (Video S1).

Solvent-Resistant Liquid Dosing System

The developed miniaturized dosing system provides full solvent compatibility for the first time. The solvent-resistant dosing (Figure 2a, Instruction S1) of sample solutions and mobile phase solvent mixtures was created in a minimalistic way with a 3D-printed syringe pump and a bought mini valve with an atomizer nozzle (Figure 1, parts 1 and 2). This was compatible with the idea of low-cost equipment because it needed no gas supply as required for conventionally used aerosol spray-on deposition.³² The liquid was pressurized and atomized for sample application and development, whereby the drop ejection was precisely controlled through a valve opening. The nozzle was positioned using two motors and respective timing belts to move the nozzle in the x-direction and the high-performance thin-layer chromatography (HPTLC) plate in the y-direction.

Figure 2.

3D print models. (a) Motorized 3D-printed syringe pump with a miniaturized valve and atomizing nozzle, enabling the precise dropwise fluid ejection and also resistance to organic solvents. (b) Multifunctional heatable plate holder for sample application and with glass cover for plate development and mini-incubation. (c) Same with dispenser trough for dosing derivatization reagents or bioassay suspensions either with a vertical slit covered with a glass rod or with a fixed glass rod insert.

The noncaptive linear actuator of the syringe pump turns a lead screw transforming the rotatory movement of the motor into a linear movement of the 3D-printed syringe push. Inside the syringe push, a force sensor was placed, measuring the force necessary to press the syringe. The force information is sent to the software, where the respective pressure is calculated. In response, the software activates/deactivates the pump according to the desired pressure. If the syringe push triggers the optical endstop, the machine stops. This is a safety precaution, also used to calibrate the position of the syringe push. When the syringe pump has generated the set pressure, the valve is opened for a short time, ejecting a drop. The software also controls the opening and closing of the valve. The dispensed volume depends not only on the time the valve is opened but also on the reservoir pressure, nozzle geometry, liquid density, and viscosity. An approximation of the dispensed volume is calculated internally to achieve application amounts comparable to the conventional instruments. Among the different nozzles tested (Table S3), the dropwise ejection of nozzle NØ 0.05 mm was not robust, whereas nozzle NØ 0.08 mm was suited for the sample application but not for development. The atomizing nozzle AL 67000 with a minimal achievable volume of 0.02 μL per ejection (calculated in Table S3) provided the best results for sharp sample zones during application and stable mobile phase flow during development. Hence, it was possible to use the same atomizing nozzle for both sample application and development. For application, the sample was ejected dropwise at predefined positions on the plate to generate a sharp band, area, or spot. In contrast, the mobile phase was ejected as a continuous forward-backward line along the 10 cm bottom plate edge (alternatively in the plate middle for antiparallel development for high-throughput analyses).

Heatable Multifunctional Plate Holder

The newly designed 3D-printed plate holder was not only heat and solvent-resistant but also multifunctional (Figures 2b,c and S2). For the first time, the same plate holder supported seven functionalities. It served as a carrier for the sample application, enabled mobile phase dosing on the plate overhang for homogeneous horizontal flow, provided plate heating, allowed dosage of bioassay suspensions and chemical derivatization reagents, and served as an incubation chamber for the microorganisms of the bioassay (mini-incubator), and as support for image acquisition. The integrated heating of the plate during sample application resulted in sharper bands as the liquid evaporated faster, which is particularly advantageous for aqueous sample solutions. The plate holder was 3D-printed with the more heat-resistant polycarbonate blend carbon fiber (Table S2) since the integrated heating deformed plate holders made of polylactic acid (started at 50 °C onward) and also of more heat-resistant polymer materials, such as polyethylene terephthalate glycol, high-temperature polylactic acid, and acrylonitrile butadiene styrene. Inside the two-parted multifunctional plate holder, a polyimide thermal foil and a temperature sensor were installed (Instruction S1). They were connected to the Arduino and controlled with the Gcode M140/M190, commonly used for heatbed heating in 3D printers. For horizontal development, the magnetic closure of the plate holder was removed so that the plate (the part below the applied samples) was hanging out of the plate holder, accepting the mobile phase dosed as forward-backward-line as mentioned. A tight glass cover was attached to reduce excessive solvent evaporation during separation, which would dry out the mobile phase solvent flow.

For the first time, the homogeneous dosage of bioassay suspensions or media as well as of reagents for derivatization or layer modification was successfully solved in a very minimalistic way. Therefore, a 3D-printed dispenser trough (Figure 2c) was attached to the x-guidance rods by clicking into place and thus positioned on the plate holder frame. In this position, the dispenser trough was filled with 1.5 mL liquid, sufficient to cover the entire 10 cm × 10 cm plate (depending on the layer thickness) by driving the plate underneath it. When moving the plate, the liquid was automatedly and homogeneously dosed over the adsorbent plate through a slit. This was achieved through the cylindrical bottom of the dispenser trough, 3D-printed with a cylinder cutout set to 5.2 mm in diameter to receive a glass rod with an outer diameter of 5.0 mm. The space between the cylindrical wall and glass rod determined the slit width. This design was easily adjustable in the slit width via the selectable 3D-printed cylinder diameter cutout. This dispenser trough with a fixed glass rod insert was ideal for carrying the chemical solution for derivatization or aqueous suspension for the bioassay. The glass rod glided along the adsorbent and did not scratch the layer. This design required the lowest dosage volume and was most robust and homogeneous in the application compared to other tested models. As an alternative, a 3D-printed dispenser trough with a vertical slit of 100 μm covered by a 5 mm glass rod was used. Smaller slits than 100 μm were challenging to print with the inexpensive Prusa i3 MK3 3D printer. The inserted glass rod reduced the vertical liquid pressure column above the slit, thus uncontrolled liquid flow-out. The slit of the dispenser trough was tightly closed by a silicone mat installed in the plate holder frame (Instruction S2) when being filled with liquid.

Development

The machine automatically ejects the mobile phase onto the plate. While the valve is opened, the syringe piston is pressed, resulting in fluid ejection. The volume ejected is dependent on the syringe push movement. The speed of the nozzle movement and the syringe push speed are used to control the ejected volume per distance moved. A constant pressure inside the system is crucial to ensure a constant flow rate and, thus, even development of the samples. To guarantee the fluid flow at the opening of the valve, a starting pressure (generated before the valve opening) can be set. The precise control of the pressure during the application was not possible due to the buffer limit of the Arduino. Faster processing than eight Gcodes simultaneously caused delays, failing the development. Therefore, the mobile phase flow rate had to be fine-tuned via starting pressure, speed of nozzle movement, ejected volume, and syringe push movement. For an even and homogeneous flow, the mobile phase solvent is dosed in both directions, i.e., during the forward-backward-line movement of the nozzle. Furthermore, a feature to fine-tune the flow rate dynamically across the movement path was implemented, as mobile phase flow velocity decreases with increasing migration distance. However, again limited by the Arduino′s buffer, only five changes per path were possible for this feature. For the food dye separation up to 35 mm, this feature was not needed; however, for the water separation up to 70 mm, it was required. The fluid ejection path is set to start and end 5 mm outside the plate to minimize any possible influence of the valve opening/closing on the stability of the fluid flow. This way, the fluid stream was already fully developed when hitting the plate, ensuring an even mobile phase solvent flow.

The prototypes of the development chamber were not completely closed at the bottom side during the dosing of the mobile phase onto the plate surface. This resulted in the gas phase not being homogeneously distributed due to the gap. In addition, there was a more or less small lateral gap between the plate holder and the plate, which resulted in an additional lateral flow due to capillary forces from the side edges, and thus, uneven flow and distorted substance migration. Also, minor variations in the horizontal orientation of the plate resulted in an uneven spreading of the mobile phase driven by gravity.

The finally designed plate holder with a glass cover and plate overhang for horizontal development (Figures 2b and S2b) overcame all aforementioned difficulties. First, the plate holder featured a magnetic mechanism for guided attachment. When detached, a part of the plate was hanging out of the plate holder, on which the mobile phase (solvent or mixture) was dosed. This 5 mm plate overhang was important as it prevented the uncontrolled capillary flow of solvent along small capillary gaps between plate and plate holder and thus incoming lateral solvent flow (from the plate sides). Second, the glass cover, helpful for inspection of the development process, had a distance of 10 mm from the HPTLC layer, allowing the gas phase to form and spread depending on the volatility of the solvent. The gas phase better remained inside the development chamber by an additionally inserted glass plate strip, which closed the gap at the bottom plate side resulting from the plate overhang. It was vertically inserted at the start of the plate overhang and touched the plate surface. The mobile phase was applied on the overhanging plate part and could only enter the development chamber by capillary flow through the porous layer. This also prevented excess volume from entering and interfering with a homogeneous solvent flow. All of these improvements minimized any external influence on the gas phase formation within the closed development chamber. Note that for high-throughput analyses and simultaneous antiparallel development, the mobile phase was applied in the plate middle using two glass plates, each 100 mm × 47 mm, as cover, to leave a 6 mm slit in the middle through which the mobile phase was dosed. Third, as the accuracy of the horizontal alignment influenced the development process, four feet were designed and 3D-printed, whose height was adjusted by turning the respective screw. It was found that a slight horizontal inclination of the plate (5° angle in the direction of the application) avoided the interference of any excess liquid with the development and thus improved the spreading of the mobile phase by capillary forces. Fourth, the plate holder was 3D-printed using a more heat-resistant polycarbonate blend carbon fiber filament to allow heating up to 110 °C and, after annealing at 140 °C for 2 h, even up to 130 °C. With the integrated thermal foil and thermistor, the temperature of the plate holder was thus controllable by software inputs. Besides sample application, the generated heating was used for solvent evaporation (after development) and chemical derivatization reactions.

Capturing Bioluminescent Plate Images

So far, a special device, the BioLuminizer (www.camag.com/product/camag-bioluminizer-2), has been required to capture bioluminescent plate images. However, in the OCLab3, the new Raspberry Pi 12 MP HQ camera could be included, which offers exposure times of up to 200 s, why it could replace the BioLuminizer. The important bioluminescence detection was studied for the first time with low-cost equipment that was newly compiled for that purpose. For image recording, the plate was automatically moved back into the miniature cabinet (behind the dosage system). Complete darkness was required in the miniature cabinet to capture the bioluminescence. Hence, it had to be 3D-printed with matt black filament to avoid the least light reflection. Images of a luminescent foil mimicking a bioluminescent HPTLC plate were first recorded to evaluate the effects of different camera settings on very long exposure images (Figure S3). With an exposure time of 50 s, a high ISO value expectedly increased the brightness, while the automatic white balance (AWB) horizon or incandescent slightly increased the image quality. However, increasing the analog and digital gains did not improve the results. Experiments with metering mode, brightness, and contrast resulted in the automatic settings as optimal, while a high value for saturation increased the image brightness. Improved brightness was observed when the dynamic range compression (DRC) and the flickering mode (FM) were activated. Setting different sensor modes and metering modes did not outperform the automatic settings, and also increasing/decreasing the exposure value (EV) compensation did not improve the image quality.

In the next step, the experiences with the luminescent foil were transferred to HPTLC plates covered with a suspension of bioluminescent Gram-negative A. fischeri bacteria. The bioluminescent bacteria were homogeneously dosed onto the plate using the newly constructed and 3D-printed dispenser trough with a glass rod (Figure 2c, Instruction S2). Then, the plate holder was tightly covered with a glass plate to keep the layer wet. An exposure time (shutter speed) of 50 s and ISO 800 already revealed a bright image of the light-green luminescent bacteria. Activating DRC and FM clearly increased the image quality and thus confirmed the results obtained with the luminescent foil (Figure S4). Slight improvements were obtained with AWB horizon and a high value of saturation. With the optimal camera settings, the exposure time (shutter speed) could finally be reduced to 20 s. The bioluminescence light of the bacteria was still detectable after 10 h, which was studied for amino and silica gel phases (Figure S5), allowing to monitor long-term biological effects. Such long monitoring times of the bioluminescence have not been reported so far.

This confirms the excellent functionality of the plate holder as an incubation chamber (mini-incubator) since, via their bioluminescence, the bacteria have proven to survive on the surface for at least 10 h. In contrast, the commercial BioLuminizer compartment was not sufficiently tight to be used as an incubation chamber, and microorganisms rapidly dried out (after 0.5–1 h). The bioluminescence image was obtained in the typical natural blue-green bioluminescence of the bacteria, in contrast to the BioLuminizer delivering only the grayscale image or artificial colors (Figures 3a and S6). These added functionalities (dosing of bioassay suspension, working as a mini-incubator, and recording of bioluminescent image) have extended the system for application to the mentioned important prioritization of compounds in complex samples via their biological effect.

Figure 3.

Bioluminescent bioautogram and in vitro microtiter plate assay response, both recorded via OCLab3: (a), Plate image with methylparaben applied (0.5–4 μg/band) captured via OCLab3 (camera settings in Figure S6) versus commercial BioLuminizer, both recorded with an exposure time of 50 s. (b), 3D print model (centered well was designed to mount an RGB LED as reference for quantitative evaluation of the pixel intensities) and respective OCLab3 image of the 3D-printed microtiter plate filled with bioluminescent bacteria suspension (100 μL/well; wells filled form the shifted letters HPTLC; exposure time 20 s).

The integrated biological laboratory made it a 2LabsToGo system. The miniaturized system for recording UV/Vis/FLD/BD images after chemical derivatizations and biological assays significantly enhances the multi-imaging detection functionality. This is important for the comprehensive nontarget analysis of complex samples, which are to be analyzed as natively and comprehensively as possible to obtain effect-directed information covering the unknown bioactive compounds of a sample (most important sample part). Only the Raspberry Pi 12 MP HQ camera with a 6 mm wide-angle lens and polylactic acid filament for miniature cabinet 3D print was required to record the bioluminescence images, which sums up to € 113. In contrast, the BioLuminizer costs 224 times more (Table S1), both including software but excluding the computer periphery. Considering the low-cost Raspberry Pi computer, the cost advantage is even more significant. This equipment can be used not only for suspension cell assays but also for adherent human cell assays³³ recently shown as a planar bioassay using commercial instruments. This opens up applications in the medical field. The miniature cabinet can be printed to fit the size of microtiter plates. Instead of expensive illuminometers, this open-source bioluminescence detection system could also be used for in vitro microtiter assays, measuring the image pixel intensity in each well (Figure 3b).

Although expected soon, as long as Pi camera Python bindings are not available for a 64-bit Linux system, long-exposure images cannot be captured with the OC-Manager3 using V4L2 as camera driver (https://wiki.st.com/stm32mpu/wiki/V4L2_camera_overview), why a bash script was written to dialog-guided capture long-exposure image sequences using the raspistill command (Instruction S2).

Detection and Digital Evaluation

HPTLC–UV/Vis/FLD/BD images were captured with the newly installed high-resolution Raspberry Pi 12 MP HQ camera with a 6 mm wide-angle lens, which achieved a full plate image of 10 cm × 10 cm at a distance of only 12 cm. Capturing the plate image, the strong fish-eye effect from the wide-angle lens was automatically corrected by calibration and subsequent post-processing with OpenCV libraries (https://medium.com/@kennethjiang/calibrate-fisheye-lens-using-opencv-333b05afa0b0), which were implemented in the new software OC-Manager3. The camera was mounted onto a miniature cabinet preventing the plate from extraneous light (Instruction S1). It was 3D-printed using a matt black polylactic acid filament to avoid reflections inside. Two RGBW LED sticks mounted at the camera holder enabled white light illumination. Additionally, UV-LEDs emitting 365 and 265 nm were implemented, soldered on four PCBs laterally attached to the cabinet (Instruction S1). The camera parameters and intensity of the LEDs can be selected in the graphical user interface and are fully controlled by the open-source software OC-Manager3 allowing different settings (Instruction S3). Captured images were then transformed into videodensitograms and preprocessed using the open-source software quanTLC¹² and rTLC.¹³ The system structure was built compatible with integrating the generated results into a future cloud-based database for autonomous neural network-guided image evaluation¹⁷ for Citizen Science. Additional tools to navigate and visualize data in the cloud can be implemented to share results and compare data with the community and peers.

Software Structure Compatible with Cloud-Based Database

The new coding was based on Python′s programming language to provide the infrastructure for connection to a cloud-based database with unsupervised image analysis. This supports compatibility with Citizen Science. To control the system, an intuitive graphical user interface was built, the software OC-Manager3 (Figure S7). A detailed software manual is available (Instruction S3). This web application was developed to run on a Raspberry Pi 4 minicomputer. It receives the user input, generates the Gcodes, and sends them to the machine. Its frontend or visual interface was written in HTML, CSS, and JavaScript programming languages. For the backend, Python and Django were used. The data are already saved in a local database but can be stored in the cloud in the future.

If the user selects a button (e.g., Start), an HTTP request is generated and sent to the backend. The data are verified, validated, formatted, and saved. The data are then used to create different Gcodes to be sent to the Arduino. The Gcodes are managed by the firmware OC-Control, based on the open-source 3D printer firmware Marlin v2.0 and changed to fit our needs. It was written in the programming language C++. OC-Control running on the Arduino receives the Gcodes and translates them into actuation of the mechanical parts. For example, the Gcode G1X5Y10 activates the stepper motors to position the plate and nozzle at the relative coordinates x = 5 and y = 10. If a Gcode order is executed, OC-Control will answer the backend (Figure S7). The control of the pump, force sensor, and valve is made possible by implementing multiple new Gcodes to Marlin. The Gcode G95 retrieves the fluid pressure off of the force sensor. With the G97 Gcode, the fluid pressure can be set. G98 allows the opening of the valve for a specific time. G41 and G40 are used to toggle the valve between the two states open and closed. The user can change all relevant parameters (speed, pressure, volume, band positioning, etc.) for sample application and development, save the method in a database, and load it or export it to a CSV file.

The software calculates the respective number of fluid ejections to reach the volume and the positioning of the ejections. The input is visualized in a graph that represents the application on the plate. The necessary Gcodes to perform the task (e.g., sample application) are automatically created and sent to the firmware. Multiple routines can be executed by a button click. For example, the system can be rinsed with an adjustable amount of fluid. To place the syringe, the syringe push can be moved. Moreover, a user interface was implemented to set fan speed exhausting the OCLab3 and temperature of the heating layer.

Proof of Concept

A mixture of six water-soluble food dyes was analyzed using the new small-scale system versus commercial instrumentation. Compared to the aerosol-driven spray-on application via ATS 4, the OCLab3 did not lead to plate erosion. The sharpness of the applied bands was comparable (Figure S8). Based on the software-internal volume calculation of the OCLab3, comparable amounts were applied. Both were separated with the mobile phase mixture ethyl acetate–methanol–water–acetic acid, 65:23:11:1. This composition was not the most uncomplicated mixture to achieve a good separation. It was very challenging, containing a small amount of acid and a wide range of solvents with different volumes and volatility. After the image recording at white light illumination, the resulting OCLab3 HPTLC–Vis chromatogram showed a uniform mobile phase flow and thus straightforward zone migration, also on the edge tracks (Figure 4). In contrast, the result obtained by the 55 times more expensive commercial instrumentation (ATS 4 and HDC) showed an edge effect for the outer tracks (distortion evident as inside bowed red dye at highest hRF value). The development in the OCLab3 took 4.5 min, whereas it took 8.5 min in the HDC. Digital videodensitograms are generated via the open-source software quanTLC.¹⁴ Exemplarily, track no. 20 in the OCLab3 chromatogram was generated, showing the red, green, blue, and grayscale pixel intensities for each position. Peak heights/areas were automatically integrated, and the method precision was calculated via the relative standard deviation of the blue dye E 131 in the red channel of the videodensitogram for the first five tracks having the same volume applied (%RSD, n = 5). For the OCLab3, the mean peak height/area was 0.085/0.926 AU with a relative standard deviation of 4.1/5.8%, whereas it was 0.135/1.270 AU with a relative standard deviation of 7.0/3.6% for the ATS 4 and HDC. This instrumental comparison showed that the OCLab3 achieved comparable image results and method precisions.

Figure 4.

Proof of concept. (a) HPTLC–vis chromatograms showing the food dye mixture (E 127, E 105, E 131, E 126, E 102, and E 124 in descending order) analyzed via OCLab3 in comparison to commercial instruments (ATS 4 and HDC), both images recorded by OCLab3; (b) Track 20 was exemplarily evaluated via digital videodensitometry using the open-source software quanTLC before and (c) after automated preprocessing (negative peak inversion, smoothing, baseline correction, and warping) of the different channels.

2LabsToGo Bioprofiling of Water Samples

The challenging nontarget analysis and bioprofiling of different water samples showed the full potential as a 2LabsToGo system. Nine different types of water were collected (Table S4). To be analyzed comprehensively concerning biological effects and as native as possible, the water samples were taken directly, without any enrichment or cleanup step, if required, only centrifuged or membrane-filtered. A high sample volume (50 μL) was applied as an area. Such high-volume applications of aqueous samples cannot be performed by simple TLC. After a short front-elution for focusing the applied area to a sharp start band, the water samples were separated. The plate was dried, and the bioluminescent bacteria suspension was dosed via the dispenser trough with inserted glass rod onto the chromatogram containing the separated water samples. The bioluminescence image was monitored for over 30 min. An impairment and thus darkening of the recorded blue-green bioluminescence indicated a bioactive or toxic compound zone that needed further clarification. Among the tested water samples, the tap water does typically not reveal dark zones. However, a difference between contaminated and clean water samples is evident. Three bioactive, dark zones (Figure 5, zones I–III) were observed for landfill water, rainwater, and biogas water (tracks 4, 8, and 9, respectively). Landfill water and biogas water were expected not to be clean, but zone II found in the rainwater was surprising and explained by the fact that it was collected unprotected in an open pot overnight. This zone was not UV/Vis/FLD-active and was first detected biologically in the bioautogram. Hence, if a dark zone appears in the bioautogram of a water sample, the water is not compatible with normal drinking water. The identification, structure elucidation, and assignment of a dark zone to a substance need to be performed in a laboratory, to which the HPTLC plate can easily be delivered by mail. The hyphenation of the bioautogram with online high-resolution mass spectrometry^34,35 is a straightforward tool to characterize bioactive zones.

Figure 5.

2LabsToGo analysis of water samples. Nine different water samples (Table S4) were analyzed directly (no sample preparation) showing (a) UV-absorbing compounds in the UV 265 nm images after the different steps in the chemical laboratory and (b) bioactive contaminants as dark zones in the blue-green bioluminescent image after bioassay detection in the biological laboratory, all recorded by the Raspberry Pi 12 MP HQ camera with a 6 mm wide-angle lens. Bioactive contaminant zones I–III are evident in three water samples, whereby zone II is detectable first via the bioassay. Zone III was exemplarily evaluated digitally using the open-source software quanTLC (as in Figure 4).

Conclusions

This first open-source 2LabsToGo system is considered a milestone, providing answers to important questions. According to the saying by Leonardo da Vinci “simplicity is the ultimate sophistication”, we hyphenate sample separation with biological effect detection and offer a solution to check the quality of complex samples. The analysis can be performed anywhere and is extremely useful for all laboratories with limited equipment. The effect images are worth a thousand words and understandable across languages, which supports its compatibility with Citizen Science. It is the first solvent-resistant system that incorporates biological effect detection and thus combines the chemical with the biological laboratory. Operated under intuitive control, it is used for application, separation, heating, chemical reagent application (derivatization), bioassay application, incubation, and multidetection of complex samples. Two laboratories are combined in one system in the most minimalistic way. It will inspire researchers to use and further modify the open-source system, which can be tailored to their own needs via the freely available codes for software and firmware, files for 3D print, and detailed instructions for self-mounting. The system is low-cost, portable, and small in footprint. The viral aspect is evidently the introduction of a complete recipe for building your own chromatography equipment, including the biological effect detection from readily available materials and providing innovative tools for pressing challenges to the community.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.analchem.2c02339.

• Comparative calculation of instrumentation, bill of materials, nozzle volumes, collected water samples, 2LabsToGo concept, multifunctional plate holder, long-exposure images, long-time monitoring of the bioluminescence, bioluminescent HPTLC plate images, software interaction scheme, comparison of application zones, assembly guide of OCLab3, capturing bioluminescent plate images, manual of OCManager3. Video of OCLab3 (prototype) operation is available at https://youtu.be/7stxV_-uNGM/ (PDF)

Author Contributions

L.S. developed the software and firmware, integrated the syringe pump system and performed the dye analysis. W.S. developed the biological detection, optimized the syringe pump and plate holder, and contributed parts to manuscript writing; R.G. performed water analysis. G.E.M. received funding, supervised instrumental development and experiments, and wrote the manuscript. All authors have approved the final version of the manuscript.

The authors declare no competing financial interest.

Dedication

Dedicated to Csaba Horváth, Professor of Chemical Engineering at Yale University, for the Csaba Horváth Memorial Award 2021.