Single-molecule detection and super-resolution imaging with a portable and adaptable 3D-printed microscopy platform (Brick-MIC)

Abstract

Over the past decades, single-molecule and super-resolution microscopy have advanced and represent essential tools for life science research. There is, however, a growing gap between the state of the art and what is accessible to biologists, biochemists, medical researchers, or labs with financial constraints. To bridge this gap, we introduce Brick-MIC, a versatile and affordable open-source 3D-printed microspectroscopy and imaging platform. Brick-MIC enables the integration of various fluorescence imaging techniques with single-molecule resolution within a single platform and exchange between different modalities within minutes. We here present variants of Brick-MIC that facilitate single-molecule fluorescence detection, fluorescence correlation spectroscopy, time-correlated single-photon counting and super-resolution imaging (STORM and PAINT). Detailed descriptions of the hardware and software components, as well as data analysis routines, are provided, to allow non-optics specialists to operate their own Brick-MIC with minimal effort and investments. We foresee that our affordable, flexible, and open-source Brick-MIC platform will be a valuable tool for many laboratories worldwide.

INTRODUCTION

Research in the molecular life sciences, biomedicine, and under clinical settings heavily relies on the use of light microscopy (1, 2), biophysical techniques (3, 4), and spectroscopic assays, e.g., polymerase chain reaction (5), enzyme-linked immunosorbent assay (6), DNA sequencing (7–9), and many others (10–12). There is rapid advancement of both the instrumentation and the assays to achieve better spatial and temporal resolution or higher sensitivity for various applications including virus detection (13–17). This progress, however, typically takes place in engineering and (bio)physics labs, and it is difficult to benefit from it in applied or industry research and under clinical settings (1, 18). Reasons for this can be the mere dimensions of a setup or the inability to operate it outside of controlled lab conditions, e.g., due to missing temperature control or lack of mechanical stability. Thus, many advanced techniques cannot be used in high biosafety labs, on field trips, research ships, in hospitals, doctor’s practices, and other locations outside the lab. Consequently, there is a growing gap between the possibilities of the state of the art in microscopy and spectroscopy and what is accessible to all interested users (18, 19). While there are core facilities for imaging and biophysical techniques, these remain too few, might only provide limited infrequent access, or come with the requirement for travel. All this poses fundamental limitations since many biological and medicinal studies require long iterative refinement, samples may have to be studied locally, and point-of-care applications using advanced techniques are simply not feasible (18, 19).

In recent years, different research groups and companies have started to bridge this gap by miniaturizing microscopy research platforms and reducing costs of commercial systems. Now, the available compact microscopy setups offer high spatial and temporal resolution or high sensitivity. The setups often use commercially available optomechanical components, as has been implemented in the smfBOX (20) and miCUBE (21). On one hand, despite their performance, these setups cannot be used easily outside of optical laboratories and require substantial expertise for setting them up, maintaining them, and operating them—making them less suitable for application-oriented users. On the other hand, Oxford Nanoimager’s video–based device is small, powerful, and user-friendly but is inflexible in terms of microscope modalities (22–24). Another drawback of all the aforementioned microscopes is high costs, which are well over 100,000 €. On the contrary, three-dimensional (3D) printing with plastic materials has gained popularity to replace expensive optomechanics and parts of the microscope frame. AttoBright, a user-friendly, minimalist, confocal microscope, showed that a 3D-printed setup can facilitate single-molecule detection (25). It comes, however, with substantial limitations in terms of data quality and general adaptability compared to the smfBOX or miCUBE. Another option, the “UC2,” a camera-based microscope, represents an adaptable platform designed for educational purposes, which allows the realization of various imaging modalities, including bright-field, dark-field, fluorescence microscopy (26) and, most recently, single-molecule localization microscopy (27).

To overcome these limitations, we here introduce an open-source microscopy platform called Brick-MIC, which uses a combination of a modular 3D-printed scaffold with a minimal number of optical components. It offers the possibility to realize a variety of (fluorescence) microscopy modalities, e.g., confocal and video detection, fluorescence spectroscopy assays, state-of-the-art single-molecule detection, and super-resolution optical imaging using the same platform. The scaffold of the microscope consists of four layers made from 3D-printed plastic material (Fig. 1): a sample holder, an excitation layer, a detection layer, and a base plate for the interchangeable excitation and detection layers. All technical drawings and detailed descriptions on how to build and assemble the different Brick-MIC modalities are available in this manuscript. We further provide (compiled) Python-based data acquisition and analysis software to perform experiments with different confocal modalities, making the Brick-MIC a true open-source microscopy platform. The only requirement to start your own Brick-MIC is access to a standard, low-cost 3D printer and to purchase of a minimal list of optomechanical and optical components.

Fig. 1. Overview of the adaptable Brick-MIC design.

(A) The platform uses shape-complementary parts that can be stacked (Lego-like or Japanese “poka yoke”), consisting of different layers: the sample holder, lid, excitation, detection, and base layers. The excitation and detection layers are interchangeable, allowing to establish different imaging methods with the same platform and easy exchange of components. (B) Photograph of a confocal Brick-MIC modality as used below for analysis of labeled biomolecules in free diffusion (see the “Single-particle fluorescence detection and FCS” section).

To date, we have established various distinct modalities of the platform, from which we use three as representative examples in this paper. The first is a confocal microscope for single-particle detection, fluorescence correlation spectroscopy (FCS) (28), and time-correlated single-photon counting (TCSPC) (29, 30), with which we were able to detect individual fluorescent particles, such as freely diffusing nano-sized fluorescent beads, dye molecules, and labeled biomolecules of varying sizes. Second, we established a two-color confocal microscope that we used for single-molecule Förster resonance energy transfer (smFRET) (31, 32) experiments with microsecond alternating-laser excitation (μsALEX) (33, 34). Last, we realized a camera-based microscope that allows standard wide-field or dark-field imaging and can be upgraded to an epifluorescence microscope for fluorescence and super-resolution imaging via STORM (35, 36) and DNA-PAINT (37, 38). We consider our approach a “Swiss-knife” microscope with full flexibility and hope that Brick-MIC will be used inside and outside of research laboratories in the future due to its small size and portability, high stability, and state-of-the-art performance.

RESULTS

General considerations for the design of the Brick-MIC platform

Brick-MIC was designed with the philosophy to create a user-friendly, portable and stable, cost-effective, and adaptable platform that can be used outside of optical labs under ambient light. To reduce costs and allow robust operation, we established an adaptable microscope body which is fully 3D printed. This platform is combined with optical components such as mirrors, filters, etc. from commercial suppliers to establish one microscope modality. The printing templates (https://zenodo.org/records/10441063) and a list of optical and optomechanical components are provided in the Supplementary Materials (Brick-MIC component list). Once the microscope frames for all modalities are printed, the Brick-MIC platform enables rapid exchange between distinct microscopy modalities within minutes. We demonstrate both the straightforward assembly and the rapid exchange of modalities in movies S1 to S3, respectively. The enclosed sample holder is suitable for the incorporation of microfluidic components, such as Ibidi microfluidic slides and tubing (39), and allows to fix microscope slides with magnets (movies S1 and S2). A single-axis translation stage, which is integrated into a rack and pinion system, controls the Z-axis position of the coverslip and sample, and ensures high positional stability (Fig. 1). The design was tested to generally ensure mechanical stability via an accurate fit of the different layers and vibration dampening by use of thermoplastic polyurethane, a resilient and rubber-like material, as the base layer of the microscope (Fig. 1 and movie S1). The detection layers of Brick-MIC are all equipped with essential optical components mounted onto piezo motors for convenient alignment of the setup and use of autocalibration software (see below). All modalities of Brick-MIC, presented in this paper, require low investment costs between 10,000 and 30,000 € and can be operated with software acquisition packages provided by the respective detector suppliers or with publicly available software (see below), which do not require expensive software licenses, in line with open science practices (19).

Single-particle fluorescence detection and FCS

Confocal microscopy can be used for sensitive detection of (individual) fluorescently labeled particles and molecules in free diffusion or flow, either via burst detection (40) or FCS (28). On the basis of the idea to use Brick-MIC outside of optical labs in biomedical research, clinical diagnostics, or environmental monitoring, we established a basic modality with a single continuous-wave (CW) laser excitation source and dual-channel detection using photomultiplier tubes (PMTs) in single-photon counting mode (Fig. 2, A and B). The confocal geometry of the setup is achieved using an inversely mounted parabolic collimator that couples the emitted light into an optical fiber (OF). The OF serves as a pinhole (PH) with a diameter related to its core size. The emitted light is captured by a high–numerical aperture (NA) water-immersion microscope objective (60×, NA = 1.2, Olympus–UPLSAPO60XW) and directed by the OF into an external detection box. Here, the emission is spectrally separated via a dichroic mirror (DM) into short- (green) and long-wavelength emission (red) before reaching two PMTs (Fig. 2B).

Fig. 2. Single-molecule detection and FCS with Brick-MIC.

(A) CAD model showing Brick-MIC with optical components and light paths (pink: excitation; purple: emission; green and red: spectral split emission after the dichroic mirror, DIC). (B) Optical layout: A single laser diode (pink) expanded by a telescope (biconcave lens L1, plano-convex lens L2) focuses into the sample through an objective (Obj) lens. Emission is collected and directed by mirrors on piezo-directed optical mounts (M2 and M3) through an inversely mounted reflective collimator, coupling the emission into an OF. The OF directs light to an external detection box, spectrally separated by a DM, detected by PMTs. (C) Single-molecule time trace of 100-pM dsDNA labeled with Cy3B (D) and ATTO647N (A) at a 13-bp interdye distance. Bursts recorded in donor (DD) and acceptor emission (DA) channels under continuous green excitation. (D) Zoom-in on time traces (max counts: 50 kHz) of 500-nm TetraSpeck beads in laminar flow (2.5 μl/s) with different OF core diameters acting as a PH. (E) Representative FCS curve of a 5 nM 40-mer dsDNA sample labeled with Cy3B, using a 10-μm PH core diameter; fit parameters: number of molecules N, geometry parameter κ, diffusion time τ_D, and triplet fraction T. (F) Boxplot of FCS-based average diffusion times of a 5 nM 40-mer dsDNA sample labeled with Cy3B measured with different PH sizes. (G) Boxplot of FCS-based average diffusion times of various biomacromolecules labeled with Cy3B with different masses and hydrodynamic radii, recorded using a 10-μm PH. Data acquired via NI-Card. Error bars: SDs from n > 3 repeats.

Signals of both PMTs are read out via an affordable single-photon counter (MCC-DAQ-USB-counter) (25) in combination with a Python software provided in this manuscript. The data can be used for online inspection of fluorescence time traces in binned format, e.g., for calibration purposes, or can be exported to hdf5 single-photon data (https://github.com/harripd/mcc-daq-acquisition) for further processing. Alignment of the setup can be achieved with an automated self-alignment procedure via two piezo mirrors using concentrated solutions of, e.g., 100 nM Cy3B (see Materials and Methods and movie S4). The self-alignment procedure was directly implemented into the aforementioned Python code and allows exchange of modalities within minutes (see movies S3 and S4). A notable benefit of this design is the convenient accessibility and interchangeability of the OF, serving as a PH with tunable diameter. This allows for easy modifications to the detection volume of the microscope by selecting OFs with varying core diameters, thereby offering enhanced flexibility in tailoring the parameters to specific experimental requirements.

Using this setup, we observed single fluorescently labeled nanoparticles of varying diameters of ≥100 nm (movie S2) and individual fluorophore-labeled double-stranded DNA (dsDNA) molecules. Diffusional transits of donor-acceptor–labeled dsDNA molecules (donor Cy3B, acceptor ATTO647N in 13-bp distance) were clearly visible as coinciding bursts in both detection channels (Fig. 2C). Variation of the OF core size reveals the effect of shrinking excitation volume in fluorescence time traces of fluorescent beads (Fig. 2D). Larger PH sizes, i.e., large detection volumes, show higher and longer signal periods, higher burst detection frequency, as well as increased background compared to smaller PH diameters (Fig. 2D and fig. S2).

The high sensitivity of the setup suggests that it can readily be used for FCS measurements. We first tested and quantified parameters of a 10 nM solution of 40-mer dsDNA labeled with Cy3B (Fig. 2E). For data acquisition, we used either the described MCC-DAQ-USB-counter (25) or an NI-Card as described previously by Gebhardt et al. (41). For FCS data analyses, we provide a jupyter notebook (https://github.com/PSBlmu/FCS---analysis), which uses functions of the publicly available FRET-bursts script from the Weiss lab (41, 42), for analysis of the obtained single-photon data in the hdf5 (MCC-DAQ-USB-counter) or binary format (NI-Card). With this procedure, we could extract molecular brightness B, diffusion time, triplet lifetime, and their associated amplitudes from a standard two-component fit with diffusion and triplet (Fig. 2E). The average diffusion times of dsDNA were ~100, ~150, and ~200 μs for increasing PH diameters. The use of an OF with a core size diameter of 10 μm resulted in comparable results to those obtained from a custom-built confocal microscope (41, 42), which uses a 50-μm PH (Fig. 2F and fig. S3). The brightness of Cy3B was lower in the Brick-MIC microscope (10 kHz versus 80 kHz per molecule; figs. S3 and S4) either due to a reduced PH diameter or lower quantum efficiency of the PMTs. The analysis of diffusion times of biomolecules with varying masses and hydrodynamic radii revealed that both setups correctly assess the expected trends related to molecular mass differences (except for the nonspherical dsDNA sample; Fig. 2G).

To demonstrate the flexibility of Brick-MIC, e.g., to use pulsed instead of continuous laser excitation, we exchanged the excitation layer to contain a reflective collimator as output of a fiber-coupled laser (Fig. 3, A and B and movie S3). As a proof of concept for TCSPC, the fluorescence lifetimes of Cy3B, Alexa 546, and Atto550 fluorophore–labeled dsDNA samples were determined using this modified excitation layer on top of the emission layer of the FCS modality (Fig. 2). We found lifetimes for Cy3B, Alexa 546, and Atto550 on dsDNA of 2.82 ± 0.01, 3.91 ± 0.01, and 4.15 ± 0.04 ns (figs. S5 to S7), respectively, matching values from a home-built cuvette-based setup with one avalanche photodiode (APD; figs. S5 to S7), which was described before (41). Each of the triplicate measurements was done via pulsed laser excitation at 532-nm wavelength at a rate of 20 MHz and a power of 55 μW (LDH-P-FA-530B with PDL 828 “Sepia II” controller, Picoquant) detected by a Multiharp 150 8N (Picoquant) for the PMT variant. For the APD variant, we used a supercontinuum white light laser source (NKT SuperK Extreme EXW-12, NKT Photonics) with 80 MHz and a power of 55 μW for excitation and detection by a Hydraharp 400 (Picoquant) as described before (41).

Fig. 3. Single-molecule FRET and μsALEX modality.

(A) CAD model showing all optical components including the light path (pink: excitation; purple: emission; green and red: spectral split emission after the dichroic mirror, DIC). (B) Overview of the optical layout: The modality uses an external laser coupled to the microscope through an OF. In this arrangement, the excitation beam is collimated with a reflective collimator and then focused into the sample through an objective lens. The emission is collected by the same objective lens and is further focused via an achromatic tube lens (TL) directly onto two different SPADs. Spectral separation is achieved by a DM and appropriate band-pass filters for each detector. The M3 mirror and the DIC in the emission layer are mounted onto piezo-directed mirror holders, which are used to direct the photon stream into each detector channel. (C) Time trace of a 100 pM dsDNA sample labeled with Cy3B donor (D) and ATTO647N acceptor (A) dye at a 13-bp interdye distance. Individual bursts are recorded in three different acquisition channels: DD (green) for donor excitation and donor emission; AA (red) for acceptor excitation and acceptor emission; and DA (yellow) for donor excitation and acceptor emission. (D) Representative ES histogram of the open conformation of the substrate binding protein MalE (left side). Right side: Observation of different conformational states of MalE (open, closed, and K_D conditions) for apo (no substrate), holo (100 μM maltose), and K_D (1 μM maltose). (E) Determination of accurate FRET efficiency values and corresponding distances (R_DA) using a DNA ladder for different dye combinations. Data were obtained using acquisition via NI-Card (see Materials and Methods for details).

Single-molecule FRET and ALEX

FRET and its single-pair equivalent spFRET (or smFRET) have become an established method in the prospering toolkit of integrative structural biology (42–52). With the smFRET method, it is possible to study biomacromolecules in aqueous solution at ambient temperature and to identify conformational heterogeneity and subpopulations, measure accurate distances, and characterize conformational changes (kinetic exchange rates) (32, 34, 41–47, 53–59).

As expected from the sensitivity of the setup, it was possible to use the PMT version of our microscope (shown in Fig. 2) for ratiometric determination of apparent FRET efficiency E* of single donor-acceptor pairs under continuous green excitation with PMTs (Fig. 2C). We realized, however, that the sensitivity of the red detection channel was suboptimal, and the available dynamic range of E* would be rather limited in comparison to our home-built setups. We thus further optimized the setup for this specific application. We improved the detection efficiency of the setup by use of two single-photon avalanche diodes (SPAD; model PDM 50-Micron, Micro Photon Devices) in the detection layer. These detectors serve as single-photon counters and pinholes simultaneously due to their small active detection area of 50 μm in diameter (23). As can be seen from the scheme of the optical setup (Fig. 3, A and B), the design greatly reduces the number of required optical and optomechanical components in comparison to standard home-built setups.

Using this modality in combination with μsALEX (34) of fiber-coupled green and red lasers at an alternation rate of 20 kHz (50-μs excitation periods), we observed bursts from individual freely diffusing FRET-labeled biomolecules (Fig. 3C). The molecular brightness of Cy3B in this setup was around 60 kHz and thus far superior compared to the PMT variant (10 kHz) and only slightly lower than that of a home-built setup (80 kHz; fig. S4). From these setups, we extracted photon streams relevant for ratiometric FRET determination (Fig. 3, D and E; apparent FRET efficiency E*) and apparent fraction of donor brightness (S*) for each single-molecule transit through the confocal excitation volume. As shown for MalE, the periplasmic subunit of the maltose permease (60), both conformational states (apo: ligand-free open and holo: ligand-bound closed) and the ligand affinity can be characterized with single-protein resolution (Fig. 3D and fig. S16) (55). These experiments use FRET as a qualitative indicator for conformational changes via E*, e.g., from low to high E* values, i.e., from long to short interdye distance, respectively.

We also assessed the ability of the setup to determine accurate FRET efficiencies, E, via correction of E* for all setup-dependent parameters, i.e., spectral cross-talk for donor leakage α, acceptor direct excitation δ, normalization of detection and quantum yield differences of acceptor and donor γ, normalization of excitation intensities, and absorption cross sections of acceptor and donor dye ß, to obtain interdye distances, R_DA, using established procedures (42, 43, 47). We compared the E values for an interdye base pair (bp) separation of DNA ladder samples with different distances (in five bp steps) between the donor and acceptor dyes, ranging from 8 to 33 bp. The dependence of E as a function of bp separation differs for the three dye pairs studied due to differing Förster distances (Fig. 3E). We used values of R₀ known from the literature: Cy3-ATTO647N: R₀ = 5.1 nm (56, 57); Cy3B-ATTO647N: R₀ = 6.7 nm (57–59); Cy3B-Cy5: R₀ = 7.4 nm (for the latter, see Materials and Methods). All E values, which are independent of R₀, are consistent between Brick-MIC and the corresponding experiments performed on a home-built confocal setup (figs. S8 to S15) (41, 42). Furthermore, the derived R_DA values of all three dye combinations are internally consistent, and the interdye distances derived from the different dye pairs cannot be distinguished within the error margins, except for large distances (>8 nm) that are outside of the sensitive dynamic range of the FRET approach (figs. S8 and S9).

Camera-based light microscopy, single-molecule fluorescence detection, and super-resolution imaging

All our investigations show that confocal-based single-molecule detection is possible with a minimal and cost-effective 3D-printed microscope system (Figs. 2 and 3). As a final step, we explored the potential of Brick-MIC for camera-based imaging. Figure 4 shows the developed wide-field epifluorescence modality. In this configuration, the excitation layer is linked to the light source via an OF, similar to the aforementioned μsALEX modality. We have, however, also tested the modality with a single laser pointer as was used for the confocal-based setup described in Fig. 2.

Fig. 4. Camera-based light microscopy, single-molecule fluorescence detection, and single-molecule localization microscopy super-resolution imaging.

(A) CAD model showing all optical components, as well as the light path (pink: excitation; purple: emission). (B) Overview of the optical layout: The modality uses an external laser box coupled to the microscope through an OF. In this arrangement, the excitation beam is collimated with a reflective collimator, which is expanded using a plano-concave lens (L1) and focused with a plano-convex lens (L2) into the back focal plane of the objective lens, resulting in an even illumination of the sample. The emission is collected by the same objective lens and focused via an achromatic tube lens (TL) onto a complementary metal-oxide semiconductor (CMOS) camera creating a real image of the sample. (C) dSTORM and DNA-PAINT imaging (up and down, respectively) using 2 × 95 nm and 3 × 80 nm DNA origami nanoruler structures, respectively. Representative epifluorescence image and reconstructed super-resolution single-molecule localization image (left, white line represents 1 μm), with a zoom in (right). (D) Representative cross-sectional profile of a single DNA origami showing the zoom (yellow dotted line). (E) Imaging of Radiolarians using classical contrast methods: transmission light microscopy (left), dark-field illumination (middle), and oblique illumination (right).

We then tested the microscope for localization-based super-resolution imaging using STORM (35, 36) and DNA-PAINT (37, 38). For this, we obtained DNA origami nanorulers from GATTAquant with a single 95-nm distance (STORM) and two 80-nm distances (DNA-PAINT) between dye attachment positions on the respective origami structure (54). These fluorescent structures were sparsely immobilized on bovine serum albumin (BSA)/BSA biotin–coated surfaces and provided images as shown in the left of Fig. 4C (epifluorescence; movies S5 and S6). Applying thiol-containing photoswitching buffer (35, 36) or DNA imager strands to the solution (33) allowed using the blinking emission of individual labels to construct super-resolved images with the ImageJ plugin Thunderstorm (61). We found that beads were helpful but not strictly necessary as fiducial markers to compensate for lateral drift, since lateral drift was <250 nm over time spans of 30 min (fig. S18). With the setup, both structures were resolvable, and we obtained a localization precision from isolated dyes or binding sites of ~30-nm full width at half maximum (FWHM) and 65-nm FWHM for STORM and PAINT, respectively (Fig. 4D). This localization precision was achieved from analyzing 2500 consecutive frames.

While the device was specifically tailored for fluorescence imaging, we were also able to use it for standard transmission light microscopy or contrast-enhancing techniques such as dark-field and oblique illumination, exemplified by images of Radiolarians recorded with the setup. To obtain these images, only minimal modifications of the optical components were required. A light-emitting diode desktop lamp was used as transmission light, positioned right on top of the sample holder. In addition, the objective was interchanged for a 20× air objective with a low NA = 0.4 (1-U2B225, Olympus). Dark-field imaging, as well as oblique illumination, was achieved by covering the sample holder with aluminum foil and punching small holes in it with different patterns. For dark field, a round pattern approximately 5 cm in diameter was used. For oblique illumination, a crescent moon shape was punctured in the foil approximately 3 cm away from the objective's axis (fig. S19).

DISCUSSION

As shown in this manuscript, Brick-MIC is a high-performance and cost-effective multifunctional microscopy platform that facilitates various state-of-the-art microscopy applications including single-molecule detection and super-resolution microscopy. The modular philosophy of the platform allows for development of additional modalities not shown here, e.g., total internal reflection fluorescence, fluorescence lifetime imaging microscopy, or light-sheet microscopy. This versatility might make it a valuable technology platform not only for imaging but also for a wide range of other scientific directions, e.g., flow cytometry or spectroscopy. Notably, the compact size of the platform offers advantages beyond the confines of the laboratory and that of previous works (19–22, 24–26, 62) in terms of performance and portability, as it enables fieldwork research, such as on-site water quality assays, or being deployed in restricted locations such as in high biosafety labs. The portability and small footprint of Brick-MIC make it ideal for conducting research in challenging conditions or real-world scenarios, opening various possibilities for scientific exploration.

The philosophy behind Brick-MIC was to optimize one technical modality (as shown above) while allowing flexibility to switch between different modalities through interchangeable modules. This principle allows to largely reduce the required components to build different microscopes. Particularly for the shown confocal modalities, the Brick-MIC design eliminated many components and allowed a simple construction for use of PH. In a conventional confocal setup, the latter often involves a tube lens to focus the emission beam through the PH, followed by a second lens for focusing onto the detector. This part is also delicate in terms of alignment precision making it a weak point in terms of stability. In the Brick-MIC PMT variant, the entire PH assembly was streamlined into a single continuous unit. This unit comprises two collimators interconnected by an OF, eliminating the need for alignment procedures between the components. For the APD detection variant, the entire assembly was simplified to a single lens by using the small aperture of the detectors as the PH, as also shown previously by Gambin and co-workers (25). Consequently, the Brick-MIC alignment procedure could be simplified to a single variable, i.e., optimization of the first element before the PH allowing automated alignment via piezo mirrors. Furthermore, the reduction in size significantly shortens the light path to approximately 30 cm, thereby enhancing the strength and stability of the Brick-MIC platform. As a result, despite the lower material stiffness of the printing filament polylactic acid (PLA) compared to steel (with a modulus of elasticity of around 3.5 GPa versus 200 GPa, respectively), it proves suitable for state-of-the-art applications. We hope to have demonstrated here that the modularity of Brick-MIC allows for rapid prototyping, e.g., to “just quickly test an alternative configuration,” introduction of additional parts (e.g., different sample holders) or to realize completely distinct modules within the platform without redesigning the setup completely. In that respect, we are now not only testing our platform for applications in interferometric scattering microscopy (iSCAT) microscopy (Cordes and Hartschuh labs) but also test replacing the PMT detection box simply by a fiber-coupled optical fluorescence spectrometer (Cordes lab).

We believe that the scientific community and applied users in industry or biomedicine will play a key role to further improve the technology by testing different optical components of distinct quality, e.g., low- versus high-NA objectives. This could further help to balance the requirements for an affordable setup versus performance. In addition, extensions of the platform related to temperature control, an autofocus system for video microscopy, and improved mechanical stability (by use of different 3D printing materials) could be a topic of future research and engineering of the platform. Other directions would concern the establishment of detector modules where data recording and analysis are more user-friendly, e.g., by use of an action camera or a mobile phone, where data recording and analysis are even easier than demonstrated here.

MATERIALS AND METHODS

Sample preparation

DNA sample for FCS and smFRET

Fluorophore-labeled oligonucleotides, as described in (58), were obtained from IBA (Göttingen, Germany). The DNA single strands were annealed using the following protocol: A 100-μl solution of two complementary single-stranded DNAs at a concentration of 1 μM was heated to 95°C for 4 min and then cooled down to 4°C at a rate of 1°C/min in an annealing buffer [500 mM sodium chloride, 20 mM tris-HCl, and 1 mM EDTA (pH 8)].

Expression and purification of proteins

MalE single- and double-cysteine variants as well as the SBD2 (T369C) protein (63) were expressed and purified generally following the established and published protocols (55, 64, 65). For all MalE derivatives, the T7/lac bacterial expression vector pET-20b(+) was chosen with a C-terminal His₆-Tag. For SBD2 (T369C), an araBAD-based bacterial expression system was selected including an N-terminal His₁₀-Tag extension.

Escherichia coli BL21(DE3)pLysS competent cells were transformed with the respective vector containing the DNA coding sequences for MalE and SBD2 variants. Expression cultures of the transformant cells were started in 2 liters of LB medium supplemented with carbenicillin (0.1 mg/ml), chloramphenicol (0.05 mg/ml), and 1 % d-glucose and grown at 37°C until an optical density at 600 nm of 0.6 to 0.8 was reached. Subsequently, overexpression was initiated by addition of 0.25 mM isopropyl β-d-1-thiogalactopyranoside for pET-20b(+) and using 0.2% l-arabinose for araBAD. The cells were harvested after 2 hours and resuspended in 50 ml of lysis buffer [50 mM tris-HCl (pH 8.0), 1 M KCl, 10 % glycerol, and 10 mM imidazole] supplemented with 1 mM dithiothreitol (DTT). The collected cells were subjected to a 30-min incubation at 4°C with deoxyribonuclease I (500 μg/ml), along with one tablet of EDTA-free protease inhibitor cocktail (cOmplete, Roche) per 50 ml of culture extract. In addition, 0.2 mM phenylmethylsulfonyl fluoride (PMSF) and 1 mM DTT were included. The cells were then lysed using an ultrasonic homogenizer (Digital Sonifier 250, Branson) equipped with a 5-mm-diameter microtip probe, with parameters set at 25% amplitude, a total exposure time to ultrasound of 10 min, and time lapses of 0.5 s for ON/OFF pulse switches. Coarse cell debris were removed by centrifugation at 5000g for 30 min at 4°C, followed by an ultracentrifugation step at 208,400g for 1 hour at 4°C to remove insoluble cellular components. The overexpressed proteins were purified by metal affinity chromatography using a Nickel functionalized agarose medium (Ni²⁺-Sepharose 6 Fast Flow, Cytiva), pre-equilibrated with 5 column volumes (CV) of lysis buffer. A total of 50 ml of the cell lysate supernatant was loaded onto 4 ml of the Ni²⁺-Sepharose resin suspension and incubated overnight at 4°C. Following the immobilization, the proteins were washed twice using two times 5 CV of lysis buffer supplemented with 1 mM DTT and 0.2 mM PMSF and eluted with 1 CV of elution buffer [50 mM tris-HCl (pH 8.0), 50 mM KCl, 10% glycerol, 250 mM imidazole, 0.2 mM PMSF, and 1 mM DTT]. To eliminate the excess of imidazole, the pooled protein-containing fractions were each dialyzed overnight at 4°C against 100 volumes of dialysis buffer [50 mM tris-HCl (pH 8.0) and 50 mM KCl] with the addition of 1 mM DTT, using a 10-kDa molecular-weight cutoff (MWCO) dialysis membrane tubing (SnakeSkin, Thermo Fisher Scientific). Next, the dialysis buffer was exchanged with 100 volumes of storage buffer [50 mM tris-HCl (pH 8.0), 50 mM KCl, and 50% glycerol] supplemented with 1 mM DTT, and the proteins were left to dialyze overnight at 4°C. The obtained proteins were snap-frozen in liquid nitrogen and stored at −80°C until further use. The final concentrations were determined at a microvolume Spectrophotometer (NanoPhotometer N60, IMPLEN) using molar extinction coefficients of 66,350 liters mol⁻¹ cm⁻¹ for MalE derivatives and 30,370 liters mol⁻¹ cm⁻¹ for SBD2.

Labeling and labeled protein purification

The stochastic maleimide labeling and purification followed an already established protocol (55, 64, 65). For each labeling reaction, 600 μg of protein from frozen stocks was used. His₆-tagged MalE (S352C) and MalE (T36C-S352C) as well as the His₁₀-tagged SBD2 (T369C) were incubated in labeling buffer [MalE variants: 50 mM tris-HCl (pH 7.4) and 50 mM KCl; SBD2: 50 mM tris-HCl (pH 7.6) and 150 mM NaCl] supplemented with 1 mM DTT to retain the reduced state of the introduced cysteine residues. In a first step, the proteins were immobilized by metal affinity on a Nickel functionalized agarose medium (Ni²⁺-Sepharose 6 Fast Flow, Cytiva), and subsequently, the maleimide reaction with 25 nmol of Cy3B for MalE (S352C) and SBD2 (T369C) (samples for FCS measurements) or with the combination of 25 nmol of each the Alexa Fluor 555 (Thermo Fisher Scientific) and the Alexa Fluor 647 (Thermo Fisher Scientific) for MalE (T36C-S352C; smFRET measurements) was carried out in the protein-specific labeling buffer overnight at 4°C. The labeled, resin-bound proteins were washed with 1 CV of the respective labeling buffer and eluted with 500 μl of labeling elution buffer [MalE variants: 50 mM tris-HCl (pH 8.0), 50 mM KCl, and 500 mM imidazole; SBD2: 50 mM tris-HCl (pH 7.6), 150 mM NaCl, and 500 mM imidazole]. Following the maleimide labeling, the single, Cy3B-labeled proteins were purified by size exclusion chromatography (ÄKTA pure chromatography system, Cytiva; Superdex 75 Increase 10/300 GL, Cytiva), and the MalE (T36C-S352C) was purified by anion exchange chromatography IEX (ÄKTA pure chromatography system, Cytiva; MonoQ 5/50 GL column, Cytiva).

Since the IEX purification will be topic of a forthcoming publication, we provide only a short synopsis here. For MalE (T36C-S352C), the eluate from the maleimide labeling protocol was prepared for further purification by removal of the remaining KCl and imidazole from the labeling elution buffer that could otherwise interfere with the anion exchange process. This step was done using a Sephadex G-25 medium (PD MiniTrap G-25, Cytiva). The labeled protein was then eluted in 1 ml of anion exchange sample buffer [10 mM tris-HCl (pH 7.5)]. The anion exchange column was set up with a 5-CV H₂O_dd wash, a 10–column volume equilibration with anion exchange sample buffer, 10–column volume equilibration with anion exchange elution buffer [10 mM tris-HCl (pH 7.5) and 1 M NaCl], and a final 20–column volume equilibration with anion exchange sample buffer. The wash and all equilibrations were done with a flow rate of 1 ml/min. Labeled MalE (T36C-S352C) was loaded onto the column with 0.5 ml/min, and subsequently, the resin-bound protein was washed with 10–column volume anion exchange sample buffer and a flow rate of 1 ml/min. For the consecutive elution, a linear increase in anion exchange elution buffer ratio with a slope corresponding to 7.5 mM NaCl per column volume was chosen. The flow rate was adjusted to 0.5 ml/min. Fractions containing MalE (T36C-S352C) with both fluorophores (Alexa Fluor 555 and Alexa Fluor 647) in nearly stoichiometric amounts were selected and used for further analysis.

Sample immobilization (STORM)

For every experiment, an ibidi μ-slide eight-well glass-bottom chamber was used. A single μ-slide chamber was washed three times with 500 μl of phosphate-buffered saline (PBS) before each experiment. Then, 200 μl of a BSA-biotin solution {1 mg/ml in PBS [140 mM NaCl, 10 mM phosphate buffer, and 3 mM KCl (pH 7.4)]} with 100-nm tracking fluorescent beads (TetraSpeck, Thermo Fisher Scientific) was added to the μ-slide chamber and incubated for 10 min. The BSA-biotin was removed, and the chamber was carefully washed three times with 500 μl of PBS. Next, a 200-μl streptavidin solution (1 mg/ml in PBS) was added to the μ-slide and incubated for 10 min. The streptavidin solution was carefully removed, and the chamber was washed three times with imaging buffer (IB; PBS with 10 mM magnesium chloride). A 1- to 10-μl solution of DNA origami nanorods (Gatta-STORM 94R, Gattaquant) was diluted with 200 μl of IB and added to the μ-slide chamber. The DNA origamis were incubated until a density of 1/μm² was reached. Subsequently, the DNA origami solution was removed, and the chamber was washed three times with 500 μl of IB. The photoswitching of the fluorophores during imaging was achieved by adding 500 μl of an oxygen scavenging system buffer (63) (pyranose oxidase at 3 U/ml, catalase at a final concentration of 90 U/ml, and 40 mM glucose in PBS) mixed with 0.1% (v/v) β-mercaptoethanol.

DNA-PAINT

DNA-PAINT samples were obtained ready to use from Gattaquant (Gatta-PAINT 80RG, Gattaquant, Germany).

3D printing of the Brick-MIC platform

All models were designed and conceived using Onshape versions 1.114 to 1.172. 3D printing was carried out using PLATech filament (OLYMPfila) on an Ultimaker +2 Extended fitted with a 0.4-mm nozzle. All models were printed with an infill density of 17%, three layers for outer walls and with a layer height of 0.1 mm. The printing speed was set to 50 mm/s, and the nozzle temperature was maintained at 210°C. To prevent warping, all parts were printed with a brim and without any supports. All models are available as STEP files as Supplementary CAD files.

Data acquisition and data analysis

μFCS

All samples were studied by positioning the confocal excitation volume into a 100-μl PBS droplet with concentrations ranging from 5 to 10 nM on a coverslip passivated with BSA (1 mg/ml in PBS). The experimental setup shown in Fig. 2 used a 532-nm wavelength CW laser diode (5-mW output; CPS532, Thorlabs) as excitation light source. The beam underwent filtration through a clean-up filter (FL05532-10 ⌀12.5 mm, Thorlabs), attenuation via a continuous neutral density filter wheel (NDC-50C-2M, Thorlabs), and expansion using a telescope comprising a biconcave lens (f = −50 mm, KBC043AR.14, Newport) and a plano-convex lens (f = 150 mm, LA1433-A-ML, Thorlabs). A dichroic beam splitter with high reflectivity at 532 nm (ZT532/640rpc, Chroma, USA) separated the excitation and emission beams to and from a high-NA apo-chromatic objective (60×, NA 1.2, UPlanSAPO 60XW, Olympus, Japan). The emitted fluorescence was collected by the same objective, directed via a mirror into a piezo-directed optical mount (AG-M100N, Newport) through an inversely mounted 12-mm reflective collimator (RC12FC-P01, Thorlabs), which focused and coupled the emission beam into a multimode OF (10-μm fiber core diameter, M64L01, Thorlabs). The fiber directed the emission light into a detection box, where it was collimated with a fixed-focus collimator (F220FC-532, Thorlabs) and then spectrally split into two separate photon streams by a DM (ZT640rdc longpass, Chroma, USA). Individual photon streams were filtered with band-pass filters (for the green channel: FF01-582/75-25 ⌀25 mm, Semrock Rochester NY, USA; for the red channel: ET700/75m, Chroma) and detected by two distinct PMTs with different spectral sensitivities (for the green channel: H10682-210, Hamamatsu, Japan; for the red channel: H10682-01, Hamamatsu, Japan). The detector outputs for FCS analysis were either recorded by a NI-Card PCI-6602 (National Instruments, USA) using LabView data acquisition software from the Weiss laboratory (66) or via a counter/timer device module (USB-CTR04, Measurement Computing, USA) with custom-made acquisition software written in Python, available for download as compiled executable or editable Python code at https://github.com/harripd/mcc-daq-acquisition. We provide information on which counter module was used for measurements in the respective text and figure caption.

Data analysis was performed using a home-written Python script (41) (https://github.com/PSBlmu/FCS---analysis), where the intensity fluctuation of freely diffusing molecules F(τ) is analyzed via autocorrelation

$$G(\mathrm{\tau })=\frac{<F(t)F(t+\mathrm{\tau })>}{<F{\left(t\right)}^2>}-1$$ (1)

Here, the correlation amplitude G(τ) describes the self-similarity of the signal in time. Average fluorescence intensities at time points t and later lag times τ are used for analysis. G(τ) is analyzed with a 3D diffusion model

$$G(\mathrm{\tau })=\frac{1}{N}{\left(1+\frac{\mathrm{\tau }}{{\mathrm{\tau }}_{\mathrm{D}}}\right)}^{-1}{\left(1+\frac{\mathrm{\tau }}{{\mathrm{\tau }}_{\mathrm{D}}}·\frac{r_0^2}{z_0^2}\right)}^{-1/2}$$ (2)

Here, N represents the average number of molecules in the confocal volume, τ_D is the average diffusion time, and r₀ and z₀ define the lateral and axial radial distances of the detection volume, which define the geometry parameter $\mathrm{\kappa }=\frac{z_0}{r_0}$.

μALEX

All samples for smFRET analysis were measured in a 100-μl PBS droplet with concentrations ranging from 50 to 100 pM on a coverslip passivated with BSA (1 mg/ml in PBS). The experimental setup used alternating laser excitation (ALEX) of two diode lasers: OBIS 532-100-LS (Coherent, USA), operated at 60 μW for donor molecules at 532 nm, and OBIS 640-100-LX (Coherent, USA), operated at 25 μW for acceptor molecules at 640 nm, both in alternation mode with a 100-μs period (50 μs for each). The lasers were combined by an aspheric fiber port (PAF2S-11A, Thorlabs, USA), coupled into a polarization-maintaining single-mode fiber P3-57 488PM-FC-2 (Thorlabs, USA) and collimated (RC12APC-P01, Thorlabs, USA) before entering an epi-illuminated confocal microscope (Olympus IX71, Hamburg, Germany). Excitation and emission collection was done by the same water immersion objective (60×, NA 1.2, UPlanSAPO 60XW, Olympus, Japan), and spectral separation was achieved by a dual-edge beamsplitter ZT532/640rpc (Chroma/AHF, Germany). Fluorescence emitted from the sample was collected by the same objective and further focused via an achromatic lens (AC254-200-A, Thorlabs) directly into the SPADs (PDM 50-Micron, MPD). The small active area of the detectors (ø 50 μm) served as a PH. Before that, the photon streams were spectrally split into the donor and acceptor channels by a single-edge DM H643 LPXR (AHF, Germany). Fluorescence emission was filtered by band-pass filters directly in front of each detector: for the donor channel FF01-582/75-25 (Semrock/AHF, Germany) and ET700/75m Chroma (AHF, Germany) for the acceptor channel. The detector outputs for μsALEX were recorded by an NI-Card PCI-6602 (National Instruments, USA) using LabView data acquisition software from the Weiss laboratory (66).

Data analysis was performed using a home-written software package, as described previously (65). Single bursts were identified first using all-photon burst search with a threshold for burst start/stop of 15 photons (40), a time window of 500 μs, and a minimum total photon number of 150 within the burst. On the basis of these data, donor leakage (α) and direct acceptor excitation (δ) were determined as mean values from a 1D fit of background-corrected donor-only E* and acceptor-only S* distributions. Then, a dual-channel burst search was performed with similar parameters to determine excitation flux (β) and detection efficiency and quantum yields (γ) (42). E-histograms of double-labeled FRET species were generally extracted by selecting 0.3 < S < 0.7. E-histograms were fitted with a Gaussian function according to $\frac{1}{\mathrm{\sigma }\sqrt{2\mathrm{\pi }}}{e}^{-\frac{{(E-\mathrm{\mu })}^2}{2{\mathrm{\sigma }}^2}}$ , where E represents the measured FRET efficiency for every detected molecule, μ is the mean, and σ is the SD.

For FRET efficiency to interdye distance conversion, the Förster equation was used

$$R_{\text{DA}}=R_0\sqrt[6]{(1-E)/E}$$ (3)

The Förster radius R₀ is given by the following equation (67)

$$R_0^6=\frac{9\text{ln}(10)}{128{\mathrm{\pi }}^5 N_{\mathrm{A}}}\frac{{\mathrm{\kappa }}^2}{n^4}{Q}_D\frac{{\displaystyle {\int }_0^{\infty }}{F}_{\mathrm{D}}(\mathrm{\lambda }){\mathrm{\epsilon }}_{\mathrm{A}}(\mathrm{\lambda }){\mathrm{\lambda }}^4\mathrm{d\lambda }}{{\displaystyle {\int }_0^{\infty }}{F}_{\mathrm{D}}(\mathrm{\lambda })\mathrm{d\lambda }}$$ (4)

where N_A is the Avogadro constant, κ² is the dipole orientation factor, n is the averaged refractive index of the medium, Q_D is the donor quantum yield, F_D is the donor emission spectrum, and ε_A is the acceptor absorbance spectrum. We used values of R₀ known from the literature: Cy3-ATTO647N: R₀ = 5.1 nm (56, 57); Cy3B-ATTO647N: R₀ = 6.7 nm (57–59). For Cy3B-Cy5, we used an unpublished approach to determine R₀ = 7.4 nm by least-square fitting of the data (E values) to the known interdye distances of donor and acceptor on the dsDNA (58) using Eq. 3.

Time-correlated single-photon counting

The setup for fluorescence lifetime decay measurements combines the excitation layer of the ALEX modality with the emission layer of the FCS modality (see above and movie S3). The configuration used a 532-nm wavelength fiber-coupled pulsed laser operating at a rate of 20 MHz and a power of 55 μW (LDH-P-FA-530B with PDL 828 “Sepia II” controller, Picoquant, Germany), detected by PMTs connected to a Multiharp 150 8N (Picoquant, Germany). All samples were measured in a 50-μl PBS droplet with concentrations of 10 nM on a coverslip. The Cy3B and Alexa 546 fluorophore-labeled dsDNA samples were measured for 10 min, and the Atto 550 sample was measured for 60 min. The instrument response function (IRF) was acquired by laser excitation of a 50-μl PBS droplet for 60 min. Lifetime decays were determined using the SymPhoTime 64 analysis program and fitted using a single Exponential Tailfit model (n = 1). For the Cy3B and Alexa 546 samples, a threshold window of 21.79 ns was used for the fitting, and for the Atto 550 samples, a window of 16.15 ns was used (see fig. S6).

μEpi

The setup for wide-field imaging uses an external fiber-coupled laser box (READY BeamTM ind 2 1007773, Fisba, Switzerland). A 640-nm wavelength continuous-wave excitation laser was used which provides an output of 30 mW (measured after the objective). The beam was collimated through a parabolic mirror (RC04APC-P01, Thorlabs), passes through a clean-up filter (ZET 635/10 ⌀25 mm, Chroma), and was expanded with a plano-concave lens (f = −25 mm, LC1054-ML - Ø1/2, Thorlabs). A plano-convex lens (f = 60 mm, LA1134-A-ML - Ø1, Thorlabs) focuses the beam into the back focal plane of the objective. A dichroic beam splitter with high reflectivity at 640 nm (ZT532/640rpc, Chroma, USA) separates excitation and emission beams into and from a high-NA apo-chromatic oil immersion objective (60×, NA 1.35, UPlanSApo60XO, Olympus, Japan). Fluorescence emitted from the sample was collected by the same objective and was further focused via an achromatic lens (AC254-150-A-ML, Thorlabs), projecting a real image onto the chip of a complementary metal-oxide semiconductor camera (U3-30C0CP-M-GL rev.2.2, IDS) (68). The photon stream is further filtered with a band-pass filter (ET700/75m Chroma, AHF, Germany) before reaching the camera sensor.

The photoswitching of the fluorophores was achieved by adding 500 μl of an oxygen scavenging system buffer (63) (pyranose oxidase at 3 U/ml, catalase at a final concentration of 90 U/ml, and 40 mM glucose in PBS) mixed with 0.1% (v/v) β-mercaptoethanol. The laser power used for imaging was 30 mW, which corresponds to approximately 37 kW/cm² with an illuminated area of 80 μm by 80 μm (fig. S17). Diffraction-limited recordings were acquired using the original software that came with the cameras. All recordings were done at 10 frames per second, with an exposure time of 100 ms and the analog gain set at maximum.

Super-resolution image reconstruction was performed using the ImageJ plug-in Thunderstorm (61). Each localization was filtered using a B-Spline wavelet filter as described. The local maxima method was used for the approximate localization of the molecules. Subpixel localization was achieved by fitting an integrated Gaussian model using a weighted least squares method with a three-pixel fitting radius. The super-resolution image was rendered with a pixel size of 12 nm × 12 nm. For STORM, drift correction was implemented using the fiducial marker algorithm with fluorescent TetraSpeck beads (100 nm, Thermo Fisher Scientific) on the sample, and lateral drift was assessed using tracking of bead positions over ~30 min (fig. S18). The maximal search tracking distance was set at 20 nm, and the minimum visibility ratio was maintained at 0.9 per frame. For PAINT, drift correction was performed using a cross-correlation algorithm, correlating the positioning of all blinking events of each molecule in the movie over time.

Supplementary Materials

This PDF file includes:

Figs. S1 to S19

Tables S1 to S16