Fig 1.
LUMoS spectral unmixing algorithm.
Fig 2.
(A) Synthetic emission spectra of 8 fluorophores. Bandwidth of the 4 detection channels were marked in shaded areas. For simplicity, the emission spectra of synthetic fluorophores were assumed to be the same Weibull distribution with the peaks of all fluorophores evenly distributed between 420nm and 685nm. (B) Spectral signatures of the 8 synthetic fluorophores in A. The intensity of each fluorophore was measured as the integrated area under the spectral curves in A. (C) Synthetic 2PLSM images based on the emission spectra in A. The ground truth image shows the 8 synthetic fluorophore expressing structures. Ch1-Ch4 images were the raw images from the 4 detection channels with an SNR of 10. Each fluorophore was synthetized to be expressed in a narrow band either vertically or horizontally. 7 of the 8 bands had the same area, while 1 small band (furthest red fluorophore) has an area 1/5th (cluster size ratio 0.2) of the rest.
Fig 3.
LUMoS unmixing of BPAE cells with channels bleed-through.
(A) BPAE cell stained with DAPI in nuclei, AF488 in actin, and MitoTracker Red in mitochondria, and imaged with 2PLSM. Images shown were 2D maximum intensity projections of 3D z-stacks. The green channel (Ch2) had a mix of actin and nuclei with the DAPI signals bleeding into the AF488. (B) LUMoS unmixing results of the mixed images in A. Only Ch1 and Ch2 images were used for separation. Note the clear separation of the nuclei from the green channel after unmixing. Background pixels were removed. (C) The theoretical emission spectra of DAPI, AF488 and MitoTracker Red. The filter bandwidths were plotted as shaded areas. Note the long tail of the DAPI spectrum blending into the green channel. (D) The relative intensities of the LUMoS unmixed DAPI and AF488 pixel clusters detected by the green and blue channels. Background was separated as an additional cluster with relatively low intensity in both channels.
Fig 4.
LUMoS unmixing of 5-color beads mixed in 4 detection channels.
(A) Mixed beads stained with Light Yellow (LY), FITC, PE, Purple, and APC imaged with 2PLSM. LY and Purple are special dyes by Spherotech. It was unable to resolve 5 colors with 4 detectors, and there were also beads shown in more than one channels. Two examples are pointed out by white arrows. (B) The images shown in A after processing through LUMoS. The first five images show the fluorescence of the five different beads separated from the 4 detection channels by the LUMoS and the last image is the composite showing all five beads as clearly separated objects. (C) Theoretical emission spectra of the 5 fluorophores. LY and Purple spectra were obtained from Spherotech, and FITC, PE and APC were obtained from online spectra-viewer. There were significant overlaps of all the 5 fluorophores. (D) The relative intensity of the pixels of each separated fluorophore in the 4 channels. Each fluorophore was represented with a unique spectral signature. Background pixels formed one additional cluster with low pixel intensities in all the channels.
Fig 5.
LUMoS unmixing of the colorful cell expressing 6 colors.
(A) Schematic of the Colorful Cell expressing BFP in nucleus, Cerulean in cell membrane, AzamiGreen in mitochondria, Citrine in Golgi bodies, mCherry in endoplasmic reticulum, and iRFP 670 in peroxisomes. (B) The 2PLSM system excitation and emission setups for imaging the Colorful Cell. 840nm and 1050nm sequential scan was conducted for the green channel (Ch2). Cerulean, AzamiGreen, and Citrine all emit significantly in the green channel. Cerulean can be excited well at both 840nm and 1050nm. AzamiGreen had more excitation at 1050nm, while Citrine excited better at 840nm. (C) The raw 2PLSM images of the Colorful Cell in the 4 channels with 2 excitation wavelengths for the green channel (Ch2). All the fluorophores were mixed in the detection channels which made it difficult to reveal individual organelles. Images were maximum intensity projections of 3D z-stacks. (D) LUMoS separation results of the images in C. 6 distinct organelles were separated into individual images and a composite image of all 6 colors is shown on the bottom. Signals from background pixels were removed. (E) The relative intensities of each separated fluorophore by LUMoS in the detection channels.
Fig 6.
LUMoS unmixing for colocalization analysis and autofluorescence removal.
(A) The raw 2-channel 2PLSM images of T cells expressing Cerulean, YFP, or colocalized Cerulean and YFP. APCs are larger cells (pointed out by white arrows) than T cells (pointed out by red arrows), and APCs are non-labeled but autofluorescent. The fluorophores were concentrated at the conjugation sites between T cells and APCs. The images were z-projections of slices 6 to 17 of 3D z-stack images (S3 Movie). The left and right cells were imaged by two acquisitions and stitched, but with the same imaging conditions. (B) LUMoS separation results of the images in A. Autofluorescence and colocalization were split into separate channels while keeping pure Cerulean and YFP signals in their own channels. Signals from background pixels were separated and removed (S3D Fig). (C) The spectral signatures of each structures produced by LUMoS. Background and autofluorescence (AutoF) were identified as additional pixel groups with distinct signatures. Colocalization (Coloc) spots were separated out due to its different spectral signature from the Cerulean-only and YFP-only groups.
Fig 7.
Simulation tests of the performance of LUMoS.
Left, the performance of LUMoS with unbalanced structure size. The number of fluorophores was fixed at 8 and SNR at 10. Middle, the performance of LUMoS with increasing number of fluorophores. The cluster size ratio was fixed at 0.2 and SNR at 10. Right, the performance of LUMoS with SNR varying. The cluster size ratio was fixed at 0.2 and number of fluorophores at 8. Results of 10 simulations were averaged to obtain all the final results.