Spectral scanning and fluorescence lifetime imaging microscopy (FLIM) enable separation and characterization of C. elegans autofluorescence in the cuticle and gut

The model organism, Caenorhabditis elegans (C. elegans), is a transparent nematode that is amenable to microscopy and study through live imaging. C. elegans imaging often uses fluorescence to analyze promoter reporters, fusion proteins or dyes that label subcellular structures (Corsi et al., 2015; El Mouridi et al., 2022; Mendoza et al., 2024; Yemini et al., 2021). Fluorescence imaging of C. elegans is central to diverse research questions from developmental biology to behavioral neurobiology (Bao et al., 2006; Chung et al., 2013; Tian et al., 2009). However, fluorescence imaging in C. elegans must contend with autofluorescence emitted from tissues and materials, such as a protective cuticle and intestinal lysosome-related organelles (gut granules) (Hermann et al., 2005; Pincus et al., 2016; Teuscher and Ewald, 2018). This was observed as early as the first account of C. elegans expressing GFP, where autofluorescence was noted to obscure the GFP signal (Chalfie et al., 1994). Methods that overcome autofluorescence in C. elegans will remove barriers to fluorescence imaging in live animals. This is especially true in areas and tissues where autofluorescence is particularly strong, such as the gut and cuticle (Heppert et al., 2016; Komura et al., 2021; Pincus et al., 2016).

Spectral approaches are a common way to overcome C. elegans autofluorescence. For example, carefully chosen bandpass filters can partially separate autofluorescence emission from GFP emission in the gut (Morris et al., 2018), intensity-based autofluorescence correction can improve the GFP signal to noise ratio in the developing embryo (Rodrigues et al., 2022), and spectral unmixing can separate fluorescent protein emission from autofluorescence (Jones and Ashrafi, 2009). Alternatively, one can rationally choose fluorescent proteins that have minimal spectral overlap with autofluorescence (Heppert et al., 2016; Thomas et al., 2019) or use non-genetically encoded fluorescent probes that emit in the infra-red range (Hendler-Neumark et al., 2021; Rashtchian et al., 2021). Finally, studies have used the genetic power of C. elegans to remove the source of autofluorescence by performing experiments in backgrounds that do not produce autofluorescent gut granules (Eichel et al., 2022).

An additional parameter that can differentiate spectrally similar fluorophores is fluorescence lifetime, which is the temporal delay between the arrival of an excitation photon and the generation of an emission photon. Each fluorophore has a unique fluorescence lifetime that depends upon the chemical structure of the fluorophore and the environment (i.e. solvent) that surrounds the fluorophore (as reviewed in Datta et al., 2020). Fluorescence lifetime can be imaged with fluorescence lifetime imaging microscopy (FLIM) and quantified through curve fitting or phasor analysis (Phasor-FLIM). Although curve fitting is widely accepted, this approach requires pre-existing knowledge about the decay parameters of the fluorophores that are being analyzed. In contrast, Phasor-FLIM analysis does not make any assumptions about the underlying decay parameters of fluorophores (as reviewed in Malacrida et al., 2021). Phasor-FLIM has been used to quantify NADH/NAD(P)H and FAD/FADH₂ ratios in metabolic studies (Bhattacharjee et al., 2017; Ma et al., 2016), separate spectrally similar fluorophores (Gonzalez Pisfil et al., 2022), distinguish known fluorophores from autofluorescence (Szmacinski et al., 2014), and quantify fluorescence resonance energy transfer (FRET) efficiency of fluorescent proteins (Lou et al., 2019). In C. elegans, FLIM has been used to investigate protein–protein interactions (Gallrein et al., 2021; Laine et al., 2019; Llères et al., 2017) and environmental effects on metabolic dyes (Chen et al., 2023). However, these studies do not address native autofluorescence in C. elegans, which has both biological relevance and a long history of complicating fluorescent protein quantification.

In this study, we performed a systematic analysis of the spectral and lifetime properties of C. elegans autofluorescence relative to the emission profiles of conventional fluorophores, such as GFP and mCherry. We show that dim GFP fluorescence can be reliably separated from bright cuticle autofluorescence using Phasor-FLIM. We also demonstrate that spectrally similar gut autofluorescence can be characterized in a label-free manner by capitalizing on heterogeneous lifetimes.

To determine the autofluorescence spectrum of live C. elegans, young adult animals were stimulated with four common excitation lines [405 nm (BFP/DAPI), 473 nm (GFP), 561 nm (mCherry), and 647 nm (emiRFP670/Alexa 647)] and non-overlapping 30 nm emission bins were collected across the entire visible and near infrared spectrum (Fig. 1A,B, Fig. S1). To quantify the spectral data, the mean pixel intensity for each emission bin was calculated across the entire animal and plotted as a spectral profile (Fig. 1C,D). In agreement with other findings (Heppert et al., 2016; Hermann et al., 2005), this approach revealed that 405 nm and 473 nm excitation stimulate autofluorescence with strong emission in the 450-600 nm range (Fig. 1B,D). Conversely, 561 nm excitation stimulates autofluorescence with weak emission in the 570-720 nm range, and 647 nm excitation produces little to no emission (Fig. 1B,D). To assess the variability of the spectral profiles, we collected data from five separate animals. We found that 405 nm and 473 nm excitation consistently stimulate strong emission (Fig. S2A-E), while 561 nm excitation produced weak and variable emission that ranged from barely detectable (Fig. S2A-C) to undetectable (Fig. S2D,E). Anatomically, the strongest autofluorescence was observed in the gut (Fig. 1B, asterisk) and cuticle (Fig. 1B, box and arrowhead). These results demonstrate that C. elegans produce a wide spectrum of autofluorescence that is distributed throughout the body of the animal, and the most intense emission overlaps with commonly used green fluorescent proteins and dyes.

Our spectral analysis agrees with the well-documented interference from autofluorescence in the C. elegans gut and the cuticle, which can make it difficult to quantify weak GFP signals (Chalfie et al., 1994; Hermann et al., 2005; Monici, 2005; Morris et al., 2018; Pincus et al., 2016; Teuscher and Ewald, 2018). Although the GFP and autofluorescence spectra overlap, we hypothesized that their lifetimes could be resolved, which would allow GFP intensity to be measured even in the presence of background autofluorescence. To test this possibility, we imaged animals expressing Podr-10::ODR-10::GFP, which is an odorant receptor protein that localizes to ciliated sensory neurons at the anterior end of the animal (Ryan et al., 2014; Sengupta et al., 1996). Regardless of the presence of the fluorescent transgene, we found that excitation using the 473 nm laser led to cuticle autofluorescence (Fig. 2A, magenta arrowhead). In animals with high levels of ODR-10::GFP expression, the GFP signal could be discerned over the cuticle autofluorescence (Fig. 2A, top row, yellow arrowhead). In animals whose ODR-10::GFP levels were relatively low, the cuticle autofluorescence obscured the GFP fluorescence (Fig. 2A, middle row, yellow arrowhead). No GFP fluorescence was seen in animals lacking the ODR-10::GFP transgene (Fig. 2A, bottom row).

To differentiate between cuticle autofluorescence and GFP fluorescence, we characterized each using Fast-FLIM (i.e. average photon arrival time, Fig. 2B). Regardless of expression level, the average photon arrival time of ODR-10::GFP was approximately 2.5 ns (Fig. 2B, top and middle rows), and the average photon arrival time of cuticle autofluorescence was approximately 1.3 ns (Fig. 2B, all rows). This suggested that FLIM could be used to separate these spectrally similar signals. However, Fast-FLIM has limited utility because it does not distinguish heterogeneous lifetimes within a single pixel. To more fully characterize cuticle and ODR-10::GFP fluorescence, we used Phasor-FLIM (Fig. 2C). On phasor plots, the ODR-10::GFP signal is located near the unit semi-circle at approximately 2.5 ns (based upon the 80 mHz repetition rate of our laser), which is indicative of a single, well-defined lifetime (Fig. 2C, yellow circle; Digman et al., 2008). In contrast, the cuticle signal is located in the interior of the unit semi-circle as a right shifted, tight cluster, which is indicative of shorter, heterogeneous lifetimes (Fig. 2C, magenta circle). These results suggest that cuticle autofluorescence arises from green fluorophores with complex decay profiles that can be spatially resolved in phasor space from the single component ODR-10::GFP. Indeed, when phasor masking is applied to these images, both bright and dim ODR-10::GFP signal can be faithfully ‘extracted’ from cuticle autofluorescence in live animals (Fig. 2D-F, top and middle rows). To demonstrate the biological usefulness of phasor masking, we used this process to characterize how the genetic mutation of the putative E2 ubiquitin ligase, ubc-6, affects ODR-10::GFP abundance. ubc-6 is a highly conserved eukaryotic gene that participates in ER-associated degradation (Christianson and Carvalho, 2022; Weber et al., 2016), but no previous studies have implicated it in olfactory receptor maintenance. We found that a deletion in the ubc-6 gene results in a 2.7-fold increase in the ciliary accumulation of ODR-10::GFP (average total photon counts for wild-type=6075, ubc-6 mutant=16,657, Student's t-test P-value <0.001; Fig. 2G,H).

Our results establish that Phasor-FLIM can separate problematic cuticle autofluorescence from GFP fluorescence in dim ciliated neurons located in the head of the animal. Next, we investigated whether spectral emission scanning and FLIM could be combined to differentiate between populations of gut autofluorescence in the anterior of the animal (Fig. 3A), which is known to result from a heterogeneous collection of subcellular lysosome-related gut granules (Hermann et al., 2005; Morris et al., 2018). Similar to our lower resolution spectral analysis (Fig. 1), the anterior gut produced heterogeneous emission spectra from individual granules that was most strongly stimulated with the 405 nm and 473 nm laser lines (Fig. 3B,C). To more fully characterize the heterogeneity, we applied K means clustering to spectral profiles of individual granules produced by 405 nm excitation. Specifically, we used the summed mean emission intensity across all wavelengths (i.e. brightness) as one component and the intensity weighted center of the emission peak (i.e. center of mass) as the second component. The clustering analysis revealed four robust populations (Fig. 3D,E, Fig. S3). The brightest population had a center of mass at approximately 525 nm (Fig. 3D,E, magenta) and included both isolated granules and granules that overlap with a larger, dimmer population (Fig. 3F,G, yellow). There were two additional relatively dim populations with centers of mass at approximately 495 nm and 510 nm (Fig. 3D-G, cyan and grey, respectively). These results demonstrate that high spatial resolution emission scanning can be combined with unbiased clustering approaches to phenotype spectrally distinct granules.

Because GFP fluorescence could be separated from spectrally similar cuticle autofluorescence using FLIM (Fig. 2), we were curious whether FLIM could also reveal different subpopulations of gut granules. To test this, we analyzed fluorescence lifetime in several anterior and posterior regions of the gut using a 473 nm excitation laser (Fig. 4A). The photon count (i.e. intensity) images revealed granules with a range of intensities. These included both homogenous granules with uniform intensity and granules that appeared to have multiple compartments (Fig. 4B,D, left column; Fig. S4). Intriguingly, some of these granules could be visually distinguished via Fast-FLIM (Fig. 4B,D, middle column; Fig. S4). We analyzed phasor plots to further understand the nature of the different fluorescent lifetime populations. Phasor analysis revealed three distinct subpopulations of multi-component autofluorescence (i.e. located in the interior of the phasor plot) that originated from spatially distinct gut particles (Fig. 4B, right column; magenta, a; yellow, b; and cyan, c). Generally, the magenta phasor population (Fig. 4C,,a) was composed of relatively large, low intensity granules that were sparsely distributed across the gut. In contrast, the yellow and cyan populations included both well-defined granules and diffuse regions of autofluorescence that lacked clear boundaries (Fig. 4C, b and c). In addition, we observed individual granules that could be separated into spatially distinct areas of the phasor plot (Fig. 4D, right column). Specifically, within a mixed population, some – but not all – granules could be separated into more than one lifetime (compare Fig. 4E, a-c, magenta, a; magenta and cyan, b; and cyan only, c). Collectively, these results demonstrate that spatially distinct gut granule autofluorescence can be more fully characterized in a label-free manner through a combination of high-resolution spectral and Phasor-FLIM analysis.

FLIM is advancing as a useful tool to overcome challenging microscopy problems (Datta et al., 2020) that include label-free analysis of autofluorescent cell structures and molecules (Blacker et al., 2014; Ouyang et al., 2021), biochemical characterization of the solvent surrounding known fluorophores (Llères et al., 2017), and distinguishing spectrally similar fluorophores (Scipioni et al., 2021). Here, we used FLIM to facilitate traditionally problematic quantification of dim GFP signal within sub-micron scale cell structures (i.e. sensory cilia) that are obscured by the green component of cuticle autofluorescence [Fig. 2 and as seen in Sepulveda et al. (2023); Wang et al. (2015)]. Compared to prior techniques, our FLIM method has the major benefit of not requiring the re-engineering of strains with different fluorescent reporters (Heppert et al., 2016) or purchasing an extensive array of overlapping bandpass filters (Morris et al., 2018). Although FLIM setups themselves can be costly and technically complex, as commercial systems become more common it is expected that using FLIM to isolate and quantify GFP signal will become more accessible. Moreover, because cuticle autofluorescence (Fig. 2) and gut autofluorescence (Fig. 4) exhibit complex decay profiles (i.e. they map to the interior of the phasor plot), the autofluorescence elimination approach described in this manuscript should be able to distinguish autofluorescence from any fluorescent protein that exhibits mono-exponential decay.

In addition, we have used FLIM to reveal sub-populations of autofluorescent lysosome-related organelles (gut granules) that can be separated based upon lifetime differences alone. This complements recent analytical approaches that combine Nile Red staining with two-photon FLIM to differentiate gut granules with distinct lipid populations (Chen et al., 2023). We have also used excitation/emission scanning to identify spectrally distinct subpopulations of gut granules that are uniquely excited at 405 nm. In the future, it will be important to identify how these spectrally distinct gut granules relate to those that can be distinguished via FLIM alone. However, this will require a pulsed ultraviolet (UV) laser to simultaneously excite the spectrally distinct population and perform time-correlated single photon counting, which is not presently available on commercial FLIM instruments.

C. elegans gut granules are an established model for understanding nutrient trafficking and metabolism. While many studies have focused on the endocytic pathways that underlie gut granule maturation, it is becoming clear that age and nutritional states can affect the physical, biochemical, and visual properties of gut granules (Chen et al., 2018; Chen et al., 2023; Hermann et al., 2005; Roh et al., 2012). For example, when animals are reared in excess zinc, gut granules form with a bilobed morphology (Mendoza et al., 2024; Roh et al., 2012). Because only one of the lobes consistently contains high concentrations of zinc, these granules are physiologically and spatially asymmetric (Mendoza et al., 2024). Our observation of some gut granules that contain fluorescence with more than one lifetime species is particularly reminiscent of these bilobed granules (Mendoza et al., 2024; Roh et al., 2012), though we did not rear animals on artificially high zinc concentrations.

The gut granules that we describe in this manuscript appear to represent spectrally defined categories, but they are not homogeneous with respect to representation and localization (Figs 3 and 4, Figs S3 and S4). This heterogeneity could arise from several aspects of C. elegans biology. First, because our samples were intentionally unlabeled, we did not attempt to identify different classifications of organelles. That is, it is possible that some of the granules that appear in our images represent lysosomes, endosomes, or other compartments derived from the endomembrane system. In addition, lysosome related organelles (LROs) undergo changes within developing and aging C. elegans. For example, protein markers for LROs can be detected during late embryonic and early larval stages (Hermann et al., 2005), but changes in lipid accumulation in LROs continue later, as the animals reach reproductive maturity and yolk proteins and lipids are transferred to maturing oocytes (Komura et al., 2021; Schroeder et al., 2007). Birefringence in LROs also increases as animals age (Komura et al., 2021). While we imaged animals after their final molt (from L4 larvae to adult animals), it is possible that our imaging captured granules that were in different stages of maturity. Finally, LROs are increasingly recognized as centers of metabolic regulation and metabolite storage. In particular, LROs can accumulate zinc (Roh et al., 2012), copper (Chun et al., 2017), and anthranilic acid glucosyl ester (downstream of kynurenine pathway, reviewed in Coburn and Gems, 2013). Importantly, even in animals experiencing a high metabolic input (for example, high levels of zinc), changes in gut granules labeling, size, and shape are heterogeneous (Roh et al., 2012). In our own analysis, we found some heterogeneity in spectral and FLIM profiles depending on where images were located (Figs S3 and S4). Overall, our data may be capturing the existing heterogeneity in the gut granule populations. Future experiments in animals lacking LROs, for example glo-1 mutants (Hermann et al., 2005; Rabbitts et al., 2008), could be used to parse the precise identity of the granules we have described. In addition, monitoring and/or intentionally modifying metabolic inputs could drive gut granules to more heterogeneous spectral profiles.

Our imaging data show that autofluorescence can be masked to remove signal that may interfere with fluorescence imaging. With respect to understanding the biology of endomembrane trafficking in the gut, this is important because gut granule autofluorescence complicates the imaging and analysis of particles as they mature (Rabbitts et al., 2008; Voss et al., 2020). Previously, researchers depended on specific filter sets and protocols to try to remove background autofluorescence (Teuscher and Ewald, 2018). Alternatively, lipophilic or metal-binding dyes have been effective at boosting the signal of organelles of interest (Mendoza et al., 2024; Sepulveda et al., 2023). Recently, gut granule stores of heme have been assessed in a dye-free assay using transient absorption microscopy, but this relies specifically on the chemical signature of heme (Chen et al., 2018). Our FLIM data suggest that biochemical differences within subpopulations, and even individual gut granules, could be differentiated without the need for labeling or knowledge of precise chemical differences. Collectively, our results demonstrate that high spatial resolution spectral scanning combined with Phasor-FLIM is a useful tool to overcome challenging live imaging problems in C. elegans biology.

N2 (Bristol), kyIs53 (Podr-10::ODR-10::GFP), kyIs53; ubc-7 (gk857464), kyIs53; ubc-6 (gk3799 gk5313[loxP]). C. elegans were maintained according to accepted protocols (Brenner, 1974; Meneely et al., 2019).

Animals were grown at 21.5°C on nematode growth media (NGM) spotted with OP-50 E. coli. Animals were age synchronized by dissolving gravid animals and allowing the remaining eggs to hatch on NGM plates (Porta-de-la-Riva et al., 2012). Age synchronized young adult animals were paralyzed in an 8 µl droplet of 30 mg/ml 2,3-butanedione monoxime on a glass coverslip for 10 min. A 2% agarose pad was used to hold the fully immobilized animals for imaging.

All imaging was performed on a Leica Stellaris 8 equipped with an 80 mHz pulsed white light laser that is tunable in 1 nm increments from 440-790 nm, a 405 nm diode (non-pulsed) laser, and five HyD detectors with dispersion-based spectral scanning from 410-850 nm. The microscope is equipped with a 63×1.4 NA oil objective, a 63×1.2 NA water objective, a 40×1.4 NA oil objective, 25×0.95 NA water objective, 20×0.75 NA dry objective, 10×0.4 NA dry objective. The microscope is controlled by LasX software that includes the Falcon FLIM module (including phasor analysis), Lightning deconvolution, and TauSense. For all imaging experiments, the 405 diode and white light laser were both turned on 45 min before data were collected to allow them to warm up. All laser intensities reported in this manuscript are relative – laser power at the sample was not determined.

To capture emission profiles of the entire animal, the 20×/0.75 objective lens was used with a digital zoom of 4.44 to create a tile scan of the animal with a 256.19 nm pixel size. To capture emission profiles of gut granules, the 63×/1.4 oil objective lens was used with a digital zoom of 5.26 to create single images with a pixel size of 68.65 nm. The focal plane for the emission scanning was approximately midway through the animal. Four commonly used excitation wavelengths (405 nm, 473 nm, 561 nm, and 647 nm) were used to create emission profiles in 30 nm increments from 420-780 nm (405 nm excitation), 480-780 nm (473 nm excitation), 570-780 nm (561 nm excitation), and 660-780 nm (647 nm excitation). The spectral scan information is stored in image stacks where each slice contains the intensity information for a 30 nm band of the emission profile (see Fig. S1B-D for an example of spectral image stack).

To create spectral plots of the entire animal (Fig. 1D) or of individual gut regions (Fig. 3C) the average intensity per unit area was calculated for regions of interest and plotted against the center of the respective emission band. To colorize the spectral image data, the slice corresponding to each emission band was converted to an RGB color corresponding to the average wavelength for that emission band [e.g. 435 nm (blue) for the 420-450 nm band and 645 nm (red) for the 630-660 nm band]. These RGB images were then summed to produce a fully colorized image. For example, if a region of interest had strong emission in the blue, green, and red bands, the summed colorized image would appear white, but if there was strong emission in green and red bands, the summed colorized image would appear yellow. The same steps were followed for ‘brightened colorized’ images, except the contrast was adjusted to saturate ≤0.125% of pixels before making the figure. To characterize gut granule emission, the R program Kmeans++ was used to cluster individual gut granules based upon the summed mean intensity (i.e. brightness) and the center of mass (i.e. color) of their spectral profiles (Fig. 3D-G and Fig. S3). The gut granules were manually outlined in FIJI prior to Kmeans++ clustering.

Images were acquired as Z-stacks with a 1 µm step size using a 63×/1.40 oil objective, a zoom of 4, and a resolution of 512×512 pixels, which leads to a 90 nm pixel size. The scan speed was set to 600 Hz with four-line repetitions and the 488 nm laser set to 100% power. The acquisition was conducted using LasX FALCON/FLIM, which sets the emission detector to single photon counting mode and synchronizes the electronics to operate as a time-correlated single photon counter. Photon count images represent the total number of photons collected at each pixel (i.e. intensity). Fast-FLIM images represent the average photon arrival time at each pixel. Phasor analysis was performed with the following settings: Pixel Binning: 1, Harmonic: 1, Threshold: 15 photons, Median Filter Radius: 11 pixels. After identifying the phasor space that contained the GFP signal and the autofluorescence signal, a circular phasor mask was created to encapsulate the appropriate area. After a region of the phasor plot was selected in LasX, the corresponding image pixels were exported as a mask. To mask GFP, a 50-pixel circle centered at 2.561 ns was used. To mask cuticle autofluorescence, a 30-pixel circle centered at 1.017 ns was used. The mask images were imported into ImageJ where all pixels outside of the mask were set to 0.

Images of wild-type and ubc-6 mutant animals were obtained with the following settings: Objective: 63×/1.40 oil, resolution: 512×512, zoom: 4, pixel size: 90 nm, step size: 1 µm, scan speed: 600 Hz, line repetitions: 4, laser: 488 nm excitation with 50% intensity. FLIM characterization was performed with the following settings: pixel binning: 1, harmonic: 1, threshold: 7 photons, median filter radius: 19. GFP signal was extracted as described above. The resulting GFP images were processed using a FIJI macro found here: https://github.com/heinohv/Dahlberg-Lab/blob/main/photon_measure.ijm.

The images were captured with the following settings: Objective: 63×/1.40 oil, resolution: 512×512, zoom: 5.26, pixel size: 69 nm, scan speed: 600 Hz, line repetitions: 8, laser: 473 nm excitation with 10% intensity. FLIM characterization and export was performing with the following settings: pixel binning: 2, hamonic: 1, threshold 20-100 photons, median filter radius: 11. To characterize different granules based on fluorescent lifetime, phasor plots were manually scanned to identify gut granules, or parts of gut granules, whose autofluorescence could be mapped back to discrete regions of phasor space.

ubc-6 and ubc-7 mutant strains were originally a kind gift from the Moerman Laboratory, University of British Columbia and are now publicly available at the CGC. The WWU Scientific Technical Services Optical Microscopy Core Facility maintained the Leica Stellaris 8 FALCON/FLIM microscope.