SlimVar for rapid in vivo single-molecule tracking of chromatin regulators in plants

Abstract

Epigenetic regulation occurs over many rounds of cell division in higher organisms. However, visualisation of the regulators in vivo is limited by imaging dynamic molecules deep in tissue. We report a technology-Variable-angle Slimfield microscopy (SlimVar)-that enables tracking of single fluorescent reporters to 30 µm depth through multiple Arabidopsis thaliana root tip cell layers. SlimVar uses rapid photobleaching to resolve tracked particles to molecular steps in intensity. By modifying widefield microscopy to minimise optical aberrations and robustly post-process few-photon signals, SlimVar mitigates performance losses at depth. We use SlimVar to quantify chromatin-protein assemblies in nuclei, finding that two homologous proteins key to epigenetic switching at FLOWERING LOCUS C (FLC) -cold-induced VERNALISATION INSENSITIVE3 (VIN3) and constitutively expressed VERNALISATION 5 (VRN5)-exhibit dynamic assemblies during FLC silencing. Upon cold exposure, the number of assembly molecules increases up to 100% to a median of ~20 molecules. Larger VRN5 assemblies preferentially colocalise with an FLC lacO transgenic reporter during prolonged cold and persist after return to warmth. Our findings support a hybrid model of epigenetic memory in which nucleation of histone trimethylation is assisted by dynamic protein assemblies over extended durations. SlimVar offers molecular insights into proteins expressed at physiological levels in tissues.

Fig. 1. Correlative quantification of diffusing assemblies using SlimVar.

Terms are defined in Table 1. SlimVar delivers a rapid photobleaching in image sequences at high (millisecond) framerate, over ~10 s cumulative exposure time t, to outpace molecular diffusion; followed by b robust postprocessing and quality control steps to identify foci in individual frames, and tracks across multiple (up to 20) frames, which correspond in general to assemblies of labelled molecules. c The full extent of photobleaching enables estimates of the characteristic molecular brightness (red arrows), which is narrowly distributed for a fluorescent protein. The characteristic molecular brightness is used to determine d total protein number for each region of interest and e stoichiometry for each tracked assembly near the start of the image sequence (blue arrows), as a number of molecules (red circles). These metrics are corrected for autofluorescence using unlabelled wild type then collated over a population, enabling robust estimation for average total protein number. f Periodicity analysis extracts patterns from the stoichiometry distribution to infer consistent repeat units of assemblies (dark circles). g Rapid tracking facilitates analysis of mean-square displacements to estimate individual assembly mobility. h Multicolour SlimVar assesses whether stoichiometry and diffusivity are dependent on colocalisation between different pairs of assemblies (white overlap between individual channels in green and magenta).

Fig. 2. SlimVar enhances optical contrast at greater working depths.

The optical scheme for SlimVar adapts widefield or objective-based TIRF (total internal reflection fluorescence) microscopy capable of detecting single molecules at a coverslip surface, and extends this to greater working depths. A narrow, collimated excitation beam is delivered at a steep but subcritical angle by (1) adjusting the position of a steering lens. The intersection of the focal plane and excitation beam defines a sub-micron high detection volume at the set working depth. The lateral size of this volume can be (2) adjusted using an iris or beam stop to match sample dimensions and reduce background. Aberrations, inherent to oil immersion lenses at depth, are mitigated at the set working depth using a calibration procedure. Either a test sample or an in vitro beads-in-agarose phantom may be used. This comprises a combination of adjustments to (3) an objective lens correction collar and, where necessary, (4) shifting the tube lens towards the objective (Created in BioRender. Payne-Dwyer, A. (2025) https://BioRender.com/13pyw8b). The microscope uses a single detector with a two-colour channel splitter (Cairn OptoSplit); note, the beam is not stopped down after entering the splitter and is shown here with a narrowed diameter only for clarity. In multicolour experiments, contrast is protected from channel crosstalk by (5) alternating excitation wavelengths between subsequent frames. The second pair of lenses in the detection path provides (6) additional magnification (1.2–2.2× depending on physical sensor pixel size) to ensure the point spread function (PSF) is spatially oversampled for super-resolved localisations.

Fig. 3. SlimVar resolves dynamics of VIN3 and VRN5 assemblies during cold exposure of root tips.

a Schematic of whole roots laid horizontally in media between agarose and coverslip for confocal and SlimVar microscopy. Created in BioRender. Payne-Dwyer, A. (2025) https://BioRender.com/13pyw8b. b Projected confocal z-stacks of VRN5-YFP root tips after 6 weeks of cold; acquisition time 35 s. Insets (interpolated) show VRN5 consistently localised to the nucleoplasm but not the nucleolus. Patterning of VEL proteins appeared round or lens-sh. aped (c.f. Supplementary Fig. 9), with median length 7.8 μm (interquartile range IQR: 5.7–10.3 μm, N = 571), and aspect ratio 1.16 (IQR: 1.06–2.10), comparable to nuclear reporters. c, d Airyscan images of VRN5-YFP after 2 weeks’ cold indicating heterogeneous distribution, shared scale bar 2 µm; c maximum intensity projection of three z-slices, averaged over three consecutive timepoints; d residence times estimated from the ratio between median and standard deviation of pixelwise values across three frames. Low standard deviation (cyan) indicates low displacement of foci over 200 ms, equivalent to diffusivity <0.1 µm²/s, while high standard deviation (magenta) indicates high displacement over 70 ms, or diffusivity >0.3 µm²/s. e Schematic indicating illumination and detection volumes (highlighted region and red box, respectively) and working depth. f–k SlimVar images of a VRN5-YFP root tip before vernalisation; shared scale bar 5 µm. f Brightfield for identifying and centring nuclei; g initial fluorescence frame, with nucleolus indicated (white dashes) and overlapping signals; h photobleaching transiently increases contrast, revealing distinct assemblies (mean projection of frames 4–6); i SlimVar resolves assemblies of different mobility on ms timescales, shown as distinct slow- (cyan, >60 ms residence time, diffusivity <0.4 µm²/s) and fast-moving (magenta, <20 ms residence time, diffusivity >1.4 µm²/s) objects, represented by pixelwise ratio of median and standard deviation. j Foci are detected from local maxima to super-resolved localisation precision. All sifted foci (Methods) for full sequence shown superimposed (white circles) on panel h (greyscale); k tracks, generated by linking nearby foci, indicate individual assemblies with independent estimates of stoichiometry and diffusivity. All sifted tracks from sequence shown with one vertex per timepoint (white arrows).

Fig. 4. Cold exposure causes VIN3 and VRN5 to form higher stoichiometry assemblies, but only VRN5 assemblies become more numerous.

a Distributions of integrated nuclear intensity (total number of labelled molecules per nucleus prior to correcting for autofluorescence) collated from cells imaged at working depths of 20 ± 10 µm at timepoints before, during and after vernalisation, for VIN3-GFP and VRN5-YFP: NV not vernalised, V2W two weeks of cold, V6W six weeks of cold, V6WT7 six weeks of cold followed by one week of warm conditions, V6WT14 six weeks of cold followed by two weeks of warm. The total protein number is the excess in integrated nuclear intensity above the mean autofluorescence in the negative control line, ColFRI (horizontal line). VIN3 total protein number is negligible before vernalisation (two-sided Brunner-Munzel (BM) test vs ColFRI, N = 33, p = 0.11: not significant at adjusted p < 0.01). However, VIN3-GFP increases sharply to ~28,000 ± 3700 molecules after 2 weeks cold (N = 64, p = 0.0031), and peaks at ~44,000 ± 4700 after 6 weeks cold (N = 83, p = 6 × 10⁻⁷). Following transfer to warm conditions, VIN3-GFP reduces to ~3200 ± 1600 molecules within 7 days (N = 37, p = 0.04). VRN5 levels increase during cold from ~110,000 ± 23,000 to ~190,000 ± 37,000 molecules (N = 94, p = 0.0089). b Numbers of tracks per nucleus (bin width = 2 for clarity; timepoints as in colour legend). VRN5 exhibits an initial increase (NV: 20.8 ± 1.9 up to 26.8 ± 1.6 tracks per nucleus at 2 weeks cold; BM test, N = 86, p = 0.0054) that is retained (27.0 ± 1.5 and 26.2 ± 2.6 tracks per nucleus at 6 weeks cold and 14 days post-cold respectively; N = 94, p = 0.80); c Collated distributions of stoichiometry (number of labelled molecules per assembly) of individual tracks (N tracks/biological replicates in Supplementary Table 1); nt no tracks detected. Bar, box and whiskers (panels a, c) denote median, interquartile range (IQR) and ±1.5 IQR respectively; cross: mean ± sem. Source data are provided as a Source Data file.

Fig. 5. VIN3 and VRN5 assemblies exhibit a two-molecule spacing in their stoichiometry distributions.

The number of labelled molecules in each assembly (stoichiometry) shows consistent peak-to-peak spacing via periodicity analysis of a VRN5-YFP and b VIN3-GFP across different vernalisation timepoints: NV not vernalised (yellow), V2W two weeks of cold (ochre/light green), V6W six weeks of cold (orange/dark green), V6WT14 six weeks of cold followed by two weeks of warm conditions (red). A kernel width (curve smoothing parameter) of 0.6 molecules was used corresponding to the standard deviation in the observed intensity of a single molecule at the sifting signal-to-noise threshold. Insets: Periodicity analysis - the number of molecules in this subunit can be estimated from the most common spacing between neighbouring peaks in each stoichiometry distribution. The threshold above which a null (aperiodic) distribution can be rejected is the 95^th percentile fraction of intervals (grey trace) output from simulated random stoichiometry (Methods). The most common interval is given by the modal kernel density estimate ± s.e.m. above the null threshold (VIN3-GFP: V2W, 1.9 ± 0.3; V6W, 2.2 ± 0.3. VRN5-YFP: NV, 1.9 ± 0.4; V2W, 2.2 ± 0.4; V6W, 2.0 ± 0.3; V6W + T14, 2.0 ± 0.4). The periodic unit in each of these cases is consistent only with an assembly subunit of 2 molecules of either VIN3-GFP or VRN5-YFP. Source data are provided as a Source Data file.

Fig. 6. Microscopic diffusivity of VIN3 and VRN5 assemblies decrease towards that of FLC loci during vernalisation.

a–c Diffusivity D of individual tracks estimated from mean-square displacement analysis at different vernalisation timepoints: NV not vernalised, V2W two weeks of cold, V6W six weeks of cold, V6WT14 six weeks of cold followed by two weeks of warm conditions. For total numbers of tracks, N, see Supplementary Table 1. a LacI-YFP tracks of fewer than 12 molecules (N = 142 tracks), detected from nuclei without pre-bleaching and identified as unbound LacI, and of LacI-YFP tracks of more than 12 molecules after pre-bleaching, identified as FLC candidates (N = 153); b Diffusivity of VIN3-GFP and c VRN5-YFP before, during and after vernalisation. VIN3 and VRN5 each exhibit a decrease in mobility during the latter part of vernalisation, persisting in VRN5 following return to warm conditions: (VIN3-GFP: D = 0.52 ± 0.03 to 0.41 ± 0.01 µm²s⁻¹; mean ± sem; N = 672 tracks, p = 0.0011; VRN5-YFP: 0.47 ± 0.02/0.48 ± 0.02 µm²s⁻¹ at NV/V2W to 0.38 ± 0.01/0.40 ± 0.02 µm²s⁻¹ at V6W/V6W + T14; N = 982, p = 0.0072). Horizontal lines denote diffusivity of FLC-lacO/LacI-YFP foci under the same conditions (solid line: mean value; grey area, agreement within error). Bar, box and whiskers denote median, interquartile range (IQR) and ±1.5 IQR respectively; cross: mean ± sem. Source data are provided as a Source Data file.

Fig. 7. A dynamic subset of enlarged VRN5 assemblies is present at FLC loci after long cold exposure and after return to warm.

a Screening for genomic FLC loci using a z-stack in the LacI-YFP channel, followed by two-colour alternating excitation in a single z-plane to capture b foci (dashed circles) and colocalisation events (solid circles) between FLC (yellow) and VRN5-mScI assemblies (magenta). c The mean stoichiometry of VRN5 when colocalised at FLC (magenta) exceeds that of uncolocalised VRN5 (grey) after vernalisation. Timepoints are: NV not vernalised, V2W/V6W two/six weeks of cold, V6W + T14 six weeks of cold followed by two weeks of warm conditions. Bar, box and whiskers denote median, interquartile range (IQR) and ±1.5 IQR respectively; cross: mean ± sem. The difference is negligible before vernalisation (3.5 ± 0.3 vs 3.5 ± 0.1, BM test, N = 365, p = 0.39) but appears at V6W (7.7 ± 0.6 vs 5.0 ± 0.2, N = 2867, p = 10⁻¹⁸) and is sustained for at least two weeks after return to warm (7.8 ± 0.5 vs 5.2 ± 0.2, N = 1416, p = 2 × 10⁻⁸). d Vernalisation preferentially increases the fraction of colocalised FLC loci with assemblies of 6 or more VRN5 molecules. Bars denote fractions with square-root estimates of standard error, while exact odds are shown above (for total detected FLC, see Supplementary Table 1). A shift to colocalisation with larger VRN5 assemblies occurs between two and six weeks’ cold, and remains on return to warm; two-tailed Fisher’s exact test: odds ratio OR = 9.7 (2.8–34.0, 95% CI), p = 7 × 10⁻⁴ (***) and OR = 9.4 (3.5–25.1), p = 3 × 10⁻⁵ (***) respectively. e Diffusivity of VRN5-mScI tracks depends on their colocalisation at FLC. Only colocalised VRN5 slow to match FLC diffusivity during late and post-vernalisation (0.39 ± 0.03 vs 0.16 ± 0.02, N = 309, p = 2 × 10⁻⁷, BM test); Number of tracks, N, legend and boxplots as for (c). f An illustration of the model for VEL-dependent epigenetic memory supported by the imaging results. Source data are provided as a Source Data file.