An Open, Integrated Platform for Multiplexed Bioluminescence Microscopy

Abstract

We report an imaging package to democratize all-in-one bioluminescence and fluorescence microscopy. The platform comprises three tools: PhasorViewer, a visualization suite to design experiments and identify optimal probe combinations; PhasorScope, an open-source, cost-effective microscopy framework to upgrade conventional microscopes; and PhasorAnalysis, a dedicated, user-friendly analysis pipeline to quantify phasor imaging data. We demonstrate the utility of the platform for multiplexed, simultaneous fluorescence and bioluminescence imaging with readily accessible optical reporters.

Introduction

Imaging tools provide “eyes” into biology and fuel new discoveries. − There remains a gap in our ability to see dynamic processes at the single-cell level, though, especially in heterogeneous environments. Conventional fluorescence approaches rely on excitation light that can induce photobleaching and phototoxicity, hindering many pursuits. , Bioluminescent platforms can circumvent these issues, ,− and dozens of luciferases are now available for microscopy applications. − However, their spectra are broad and overlapping, complicating multicomponent studies ,, Imaging with more than one reportercommonplace in fluorescenceis incredibly rare in bioluminescence microscopy.

Bioluminescent phasing addresses the key gap in multiplexing ability. This approach optically transforms spectral signatures into spectral phasors, providing phasor coordinates with high spectral resolution and uncompromised integration times. Bioluminescent phasor also retains, on average, twice as many photons as RGB cameras. We previously showcased the utility of a bioluminescent phasor for multicolor longitudinal imaging of both subcellular features and heterogeneous spheroids. We further resolved previously indistinguishable mixtures of optical reporters and provided universal and precise measurements of bioluminescence resonant energy transfer (BRET), a parameter of crucial importance in biosensor design.

While we demonstrated feasibility, key hurdles prevented easy adoption of bioluminescent phasor technology. The instrument relied on two cameras and, thus, a complicated calibration scheme. The initial design also did not support fluorescence and lacked user-friendly analysis software to resolve optical reporters. Furthermore, identifying good combinations of reporters for multiplexed imaging a priori was nontrivial.

Results

Here, we report a compact, user-friendly workflow that overcomes these obstacles and makes multiplexed bioluminescence and fluorescence imaging accessible to the broader community (Figure ). First, we developed PhasorViewer, a simple interactive tool for identifying combinations of bioluminescent probes for microscopic imaging. Second, we designed PhasorScope, an open-microscopy platform that can be easily integrated with conventional microscopes. A complete parts list (totaling ∼$7500 in 2025, excluding the price of the camera) and manual, along with a step-by-step video, guide the user through the assembly, alignment, and calibration. The PhasorScope build requires only ∼2 h for an experienced engineer and ∼8 h for a nonspecialist. Each subsequent use requires only alignment and calibration checks that are performed in minutes. Last, we produced PhasorAnalysis, an analysis pipeline featuring cell segmentation and a variety of common plotting and output files (e.g., spreadsheets and image files). PhasorAnalysis also features a simple user interface and cursors for real-time decisions and is accompanied by a video tutorial on the code. PhasorViewer and PhasorAnalysis are written in Google CoLab, a browser-based platform for Python programming, to maximize access.

1.

Workflow for integrated optical imaging. PhasorViewer: screenshot of Colab code (left), plot of simulated complex spectra (center), and corresponding phasor coordinates (right) for selecting optical reporters. PhasorScope: a photo of the assembly, highlighting the optical paths for unfiltered (white), sine-filtered (cyan), and cosine-filtered (purple) light. Insets show pictures of the sine and cosine filters as well as the three optical paths. PhasorAnalysis: sample output images (Intensity, Phasor, and Segmentation) and plots (pixels and single cells). Numbers and colors in the single-cell phasor plot correspond to the ones in the segmentation image.

The phasor-based implementation is particularly useful for the spectral unmixing of overlapping signals. Spectral unmixing is traditionally achieved via linear unmixing, which is typically implemented as a pixelwise fitting of a known spectral database to extract photon abundances that are then used to generate unmixed images. Phasor analysis has two main advantages: computational efficiency and resistance to noise. Both advantages are due to its implementation via Fourier transformation, which allows one-shot unmixing of the whole image while, at the same time, reducing the noise contribution. While other multicolor methods require spatially separated luminophores (e.g., expressed in different cells), this implementation can simultaneously image up to three, spatially overlapping luminophores.

To showcase the modularity of PhasorScope, we used two commercial cameras featuring different technologies (Figure A), a high-resolution sensor (Hamamatsu Orca Quest, qCMOS, 4.6 μm pixels, 4096 × 2304) and a large-pixel sensor optimized for sensitivity (Nüvü HNü 512, EMCCD, 16 μm pixels, 512 × 512). Both setups could readily detect cells expressing bioluminescent enzymes (YeNL in Figure A, NanoLuc and LumiScarlet in Figure S1). The analysis platform also readily reported similar phasor locations for each reporter. The system is thus robust and amenable to changes in components. Moreover, the spectral phasor coordinates can be directly compared across instruments and laboratories.

2.

Applications of PhasorScope. (A) PhasorScope imaging with different cameras. MDA-MB231 cells stably expressing YeNL were imaged with a Hamamatsu qCMOS camera (left) or a Nüvü EMCCD camera (right). (B) Detection of dim-emitting probes. Fluc-expressing HEK cells were treated with D-lucifeirn and imaged. (C) RNA imaging with a split luciferase reporter. For (B,C), Segmented cells (left), the intensity image (middle), the pixel-wise phasor distribution (top right), and average phasor location of the single cells (bottom right) are shown. Scale bars = 10 μm. (D) Multimodal imaging (bioluminescence, fluorescence, and brightfield) with cocultures of three cell lines. Three distinct live-cell organelle stains were also used. Top row, left to right: intensity-weighted unmixed bioluminescence image (HeLa BreakFAST Cyto-Nluc in blue, MDA-MB231 YeNL in green, A549 LumiScarlet in red), pixel-wise phasor plot of bioluminescence signal, average phasor location of the single cells for bioluminescence image. Middle row, left to right: intensity-weighted fluorescence image (nuclei in blue, lipid droplets in green, and lysosomes in red), pixel-wise phasor plot of fluorescence signal, and average phasor location of the single cells for fluorescence image. Bottom row, left to right: brightfield image of the same field of view as bioluminescence and fluorescence images. Boxplots show sum intensity distribution for nuclear (Hoechst 33342), lipid droplet (Bodipy 505/515), and lysosomal (LysoTracker DeepRed) signals, respectively, for each cell line (H= HeLa BreakFAST Cyto-Nluc, M = MDA-MB231 YeNL, A = A549 Lumiscarlet). Scale bar = 10 μm.

We further showcased PhasorScope for imaging ultradim signals. Firefly luciferase (Fluc)-expressing cells and D-luciferin are widely used for imaging in vivo, but their slow turnover rate has precluded many microscopic applications. More sensitive cameras are beginning to address the need, and we showed that one (Nüvü) was compatible with the PhasorScope. As shown in Figure B, Fluc-expressing cells were readily visible on the setup and registered phasors. Note that optical phasor transformation can result in phasors outside the unit circle, particularly in the presence of noise and narrow emission spectra. In terms of brightness, the differences in integration times to obtain images of comparable quality (e.g., ∼seconds for the enhanced NanoLanterns and ∼thousands of seconds for firefly luciferase) align with differences in brightness measured on conventional plate readers to aid experimental planning. In a second example, we imaged cells expressing a split luciferase capable of RNA detection (RNA lantern, Figure C). These and other split reporters are notoriously dimmer compared to their intact counterparts, but our work shows that they are amenable to bioluminescent phasor.

PhasorScope readily supports both fluorescence and bioluminescence imaging, a feature long desired in the imaging community. This dual-modality capability is enabled by key design features: addition of an LED laser excitation source and dichroic mirror, a light-efficient and light-tight optical path that maximizes photon collection while eliminating background light, and a single sensitive detector capable of capturing signals across both dim bioluminescence and brighter fluorescence ranges. Together, these components allow for sequential acquisition of both modalities in the same field of view. As shown in Figure D, we sequentially acquired brightfield (using an illumination lamp), bioluminescence (with luciferin application), and fluorescence (LED excitation) images of live cells in the same field of view. Three distinct cell types, each expressing a different luciferase (NanoLuc, LumiScarlet, or YeNL), were observed along with three unique organelle-targeting fluorophores. PhasorScope imaging provides cell morphology (brightfield), cell type (bioluminescence), and physiological signature (organelle dyes) at optical resolution in the same field of view. We then used PhasorAnalysis on the bioluminescence and fluorescence images to quantify the intensities of each luciferase and organelle dye. Ultimately, we were able to simultaneously quantify the properties for the three organelles in all three cell lines from a single, multimodal data set.

It is important to note the complementary temporal characteristics of the two modalities. Fluorescence imaging on our system operates with millisecond-scale acquisition times, while bioluminescence often requires minute-scale integration due to a lower photon flux. This temporal trade-off underlies the key strengths of bioluminescence: excitation-free operation, negligible phototoxicity, and suitability for longitudinal studies. PhasorScope thus combines fast fluorescence for capturing cellular dynamics with gentle bioluminescence for extended time-lapse experimentsproviding researchers flexibility to match imaging modality to biological timescale.

Conclusion

In conclusion, we developed an open-source, open-microscopy toolbox for planning (PhasorViewer), imaging (PhasorScope), and analysis (PhasorAnalysis) with uncompromised optical resolution and integration, democratizing multiplexed bioluminescence and fluorescence imaging for the research community. The platform will democratize phasor analysis, and the modular and custom setup opens doors to further improvements. As an example, regardless of the methods used for spectral encoding (e.g., RGB-like filters or sine-cosine filters), the theoretical maximum number of resolvable colors, mathematically, is equivalent to the number of acquisition channels. As we implemented 3-channel configuration (unfiltered, sine- and cosine-filtered), our multiplexing capabilities are restricted to three spatially overlapping colors. This configuration, though, provides two key differences with respect to RGB-like filtering: the increased efficiency of the sine-cosine configuration, retaining 2/3 of the incoming fluorescent light in place of the 1/3 of RGB-like configurations, ultimately doubling the efficiency while, at the same time, allowing for virtually infinite spectral resolution. The latter is due to the fact that the emission filters have continuous transmission profiles, and the phasor transformation is performed optically, making very small differences in spectral emission visible. Thus, we expect the phasor-based platform to further push the frontier of optical imaging.

Materials and Methods

General Cell Culture Methods

HEK293, MDA-MB-231, HeLa, and A549 cells (ATCC) and cells engineered to express various reporters were cultured in complete media: DMEM (Corning) containing 10% (v/v) fetal bovine serum (FBS, Life Technologies), 4.5 g/L glucose, 2 mM l-glutamine, penicillin (100 U/mL), and streptomycin (100 μg/mL, Gibco). For transient transfection experiments, HEK293 cells (5 × 10⁵) were plated 24–48 h prior to transfection in tissue culture-treated 6-well dishes (Corning). Transfections were performed with luc2-IRES-eGFP (Fluc) or Staygold-MS2–3-PP7 (RNA bait for RNA probe) using Lipofectamine 3000 according to the manufacturer’s instructions when cells were 75–80% confluent (1–2 d post plating). Cells were manipulated 24–48 h post transfection. Cells were incubated at 37 °C in a 5% CO₂ humidified chamber. Cells were serially passaged using trypsin (0.25% in HBSS, Gibco).

Viral Transductions

FLAG-tagged YeNL and LumiScarlet reporter plasmids (in pLenti backbones), described in Table , were cotransfected with the VSV-G envelope plasmid and d8.2 gag/pol helper plasmids into Lenti X-293T cells (seeded in 10 cm dishes). Transfections were performed using 1 mg/mL polyethylenimine (PEI, VWR #BT129700). A total of 1.5 μg of VSV-G, 5 μg of d8.2, and 6 μg of FLAG-pLenti plasmids were mixed in 500 μL of OptiMEM (LifeTech #1158021) and vortexed. In a separate tube, 25 μL of PEI (2 μL PEI/μg DNA) was added to 475 μL of OptiMEM and vortexed. Both solutions were incubated at room temperature for 5 min and then combined and incubated for an additional 20 min. The resulting mixture was added dropwise to the seeded Lenti X-293T cells. The following day, the cell culture medium was replaced with DMEM supplemented with 1× NaPyr, 10 mM HEPES, 1× GlutaMAX, and 1× PSG (no FBS). Two days later, the media was collected and filtered through a 0.45 μm PVDF membrane, and the viral particles were concentrated by ultracentrifugation (100,000g for 2 h at 4 °C using a SW28 rotor on a Beckman L8–80 M centrifuge) into a sucrose cushion. The concentrated virus was resuspended in cold PBS and either stored at −80 °C or used immediately on target cells (MDA-MB-231s and A549).

1. Table Reporting on the Name, Features, Vector Type, and Reference to the Original Manuscript and Addgene Number for the Constructs Used in This Work.

reporter	features	plasmid vector	reference	addgene#
NanoLuc	FLAG-Nluc-SV40-PuroR-WPRE-AmpR	pLenti 3rd gen	Yao, et al.	183045
YeNL	FLAG-Venus-Nluc-SV40-PuroR-WPRE-AmpR	pLenti 3rd gen	Yao, et al.	208849
LumiScarlet	FLAG-mScarlet-Nluc-SV40-PuroR-WPRE-AmpR	pLenti 3rd gen	Yao, et al.	208854
RNA probe	SV40NLS-HA-MCP-SmBiT114-IRES-PCP-LgBiT-FLAG-PuroR-WPRE-AmpR	pLenti 3rd gen	Halbers, et al.	226095
Staygold-M-3-P	Staygold-MS2-3-PP7-AmpR	pCDNA 3.1	Halbers, et al.	N/A
BreakFAST-cyto	BreakFAST-SV40-PuroR-WPRE-AmpR	pLenti 3rd gen	Torrey, et al.	N/A
firefly (Fluc)	luc2-IRES-eGFP-AmpR	pCDNA 3.1	Love, et al.	N/A

MDA-MB-231 Lumiscarlet, HeLa Nluc, or MDA-MB231 YeNL was seeded in 8-well imaging chambers (Cellvis) and imaged with either the Hamamatsu (ORCA-Quest qCMOS, C15550, Hamamatsu) or the Nüvü (HNü 512, 512 × 512 EMCCD) camera. For each cell line, we acquired 2 frames with 2 min integration. Calibration data were acquired as described above and processed using the PhasorAnalysis code (Figure S1).qCMOS/EMCCD comparison experiment.

RNA Probe, NanoLuc, and BreakFAST

Cell lines stably expressing the RNA probe, NanoLuc, and BreakFAST were generated via lentiviral transduction. Briefly, HEK293 packaging cells were plated in 6-well dishes and transfected as described above with viral packaging plasmids along with luciferase plasmids in the pLenti vector. Cells were transfected at 70% confluency. 24 h post transfection, media were replaced with viral induction media: DMEM (Corning) contained 4.5 g/L glucose, 2 mM l-glutamine, 20 mM HEPES buffer, and 10 mM sodium butyrate. After a 48 h incubation with viral induction media, media was collected from HEK cells and spun down at 500 xg for 5 min. Supernatant was plated directly on preplated and adhered destination lines (HEK, HeLa, and MDA-MB-231s), 8 μg/mL Polybrene (final concentration) was added, and cells were incubated for 2 d. Media was then removed, and fresh media was added containing 25 μg/mL puromycin for selection. Final puromycin concentrations were tested and scaled for individual cell lines. Once selected cells grew to 100% confluency in a T25 or T75 flask, they were subjected to FACS to select the brightest clones. Cells were seeded in 8-wells imaging chambers (CellVis) at a density of 30 000 cells/well, the day before imaging, and kept in the incubator. Prior to imaging (10 min), 3 μL of furimazine (50 mM, Promega) was added.

Fluc Probe

HEK293 cells (5 × 10⁵) were seeded in 8-well Ibidi μ-Slides and transfected with the Fluc-IRES-eGFP construct (Table ) using Lipofectamine 3000 (ThermoFisher) following manufacturer instructions. Cells were imaged 24 h post-transfection. Fluorescence illumination acquired using blue LED light source (ThorLabs, Solis-470C) of eGFP was used to localize positively expressing cells. For bioluminescence imaging data collection, 3 μL per well (total volume 300 μL) of the substrate D-Luc from the stock 10 mM (Goldbio catalog #LUCK-100) was added.

The following acquisition parameters were used: Nüvü camera objective 20× 0.75 NA, collection of 12 frames with a 5 min integration time per frame, and 750 gain. Calibration was performed by acquiring “bright” (brightfield lamp, 20 frames, 200 ms integration time), “dark for bright” (no illumination, 20 frames, 200 ms integration time), and “dark for measures” (12 frames, same integration time as the measures). Data were processed using the PhasorAnalysis Colab code.

RNA Probe Imaging

HEK293 cells (5 × 10⁵) stably expressing the split RNA MS2 and PP7 coat protein lantern (RNA lantern) were plated in 8-well Ibidi μ-Slides. After 24 h, the cells were transiently transfected with Staygold-MS2-3-PP7 (Table ) using Lipofectamine 3000 (ThermoFisher) following manufacturer instructions. 24 h after transfection cells were imaged. To carry on the measurement, we used the Phasor Scope with Nüvü camera for detection. To confirm the location of cotransfected cells we collected fluorescent measurements (LED laser 505 nm emission, emission filter 535/30, 2 frames with 25 ms integration time, gain 100). To collect the bioluminescence signal 2 frames with 2 min integration time, gain 100 was used.

Experiment for Multimodal Imaging (Fluorescence, Bioluminescence, and Brightfield)

For each experimental day, the following calibration data sets were collected:

• Dark for fluorescence (20 frames, 200 ms integration time)

• Dark for bioluminescence (2 frames, 2 min integration time)

• Bright (brightfield lamp, intensity set to have ∼16k counts in detection, acquired at 200 ms integration time, 20 frames).

• For single cell lines with a bioluminescence reporter, 3 μL per well of substrate furimazine (50 mM, Promega) was added 10 min before imaging. The focal plane was determined with brightfield illumination, acquiring 1 frame (200 ms integration time). Sequentially, we acquired a brightfield, a bioluminescence, a fluorescence, and, again, a brightfield image.

For coculture experiments, a 1:1:1 mixture of HeLa BreakFAST Cyto-Nluc, A549 Lumiscarlet, and MDA-MB231 YeNL was seeded in 8-well imaging chambers (Cellvis). The day of the experiment, culture media was replaced with media containing a mixture of organelle labeling dyes: Hoechst 33342 for DNA, BODIPY 505/515 for lipid droplets, and LysoTracker DeepRed for lysosomes). Bodipy 505/515 (Cayman Chemicals, final concentration 1 μM), LysoTracker DeepRed (ThermoFisher, final concentration 100 nM).

To calibrate the phasor positions for bioluminescence and fluorescence, we imaged single populations of HeLa BreakFAST Cyto-Nluc, A549 Lumiscarlet, and MDA-MB231 YeNL alone in bioluminescence (Figure S2), and we stained HeLa Nluc cells with single dyes (fluorescence, Figure S3). Multimodal images were analyzed as follows. Bioluminescence images were used for cell segmentation (Cellpose), and the bioluminescence and fluorescence images were unmixed by spectral phasor unmixing. The cells were classified as HeLa, A549, or MDA-MB-231 according to which bioluminescent reporter (Cyto-Nluc, Lumiscarlet, or YeNL, respectively) resulted in the highest relative fraction. For each classified cell line, the fluorescence signal of the organelle dyes was quantified as the total intensity of the unmixed signal within the cell masks.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/cbmi.5c00225.

• Figure S1-S3, PhasorScope parts list and PhasorScope Manuals (Alignment, Splitter, and Enclosure) (PDF)

¶.

L.S., B.T., and G.T. contributed equally. L.S., B.T., G.T., M.A.D., and J.A.P. conceived the project idea. L.S., B.T., G.T., L.P.H., Z.R.T., and E.B.F. performed the experiments. L.S. and B.T. designed and built the imaging setup. L.S., F.F., and C.G. wrote the codes and analyzed the raw imaging data. A.L. provided technical support on the imaging setup. L.S., B.T., G.T., L.P.H., Z.R.T., E.B.F., M.A.D., and J.A.P. analyzed data and contributed to the writing of the manuscript. All authors have given approval to the final version of the manuscript.

The authors declare no competing financial interest.