Condensed mitotic chromosome structure at nanometer resolution using PALM and EGFP- histones

Abstract

Photoactivated localization microscopy (PALM) and related fluorescent biological imaging methods are capable of providing very high spatial resolutions (up to 20 nm). Two major demands limit its widespread use on biological samples: requirements for photoactivatable/photoconvertible fluorescent molecules, which are sometimes difficult to incorporate, and high background signals from autofluorescence or fluorophores in adjacent focal planes in three-dimensional imaging which reduces PALM resolution significantly. We present here a high-resolution PALM method utilizing conventional EGFP as the photoconvertible fluorophore, improved algorithms to deal with high levels of biological background noise, and apply this to imaging higher order chromatin structure. We found that the emission wavelength of EGFP is efficiently converted from green to red when exposed to blue light in the presence of reduced riboflavin. The photon yield of red-converted EGFP using riboflavin is comparable to other bright photoconvertible fluorescent proteins that allow <20 nm resolution. We further found that image pre-processing using a combination of denoising and deconvolution of the raw PALM images substantially improved the spatial resolution of the reconstruction from noisy images. Performing PALM on Drosophila mitotic chromosomes labeled with H2AvD-EGFP, a histone H2A variant, revealed filamentous components of ∼70 nm. This is the first observation of fine chromatin filaments specific for one histone variant at a resolution approximating that of conventional electron microscope images (10-30 nm). As demonstrated by modeling and experiments on a challenging specimen, the techniques described here facilitate super-resolution fluorescent imaging with common biological samples.

Figure 1. Photoconversion of EGFP by Reduced Flavin.

(A) Summary of interactions among EGFP (chromophore is shown yellow), riboflavin (chemical structure), methionine (Met), molecular oxygen (O²), and blue light (488 nm wavelength). Photoreduction of riboflavin results in acquisition of two hydrogens (red) with a bond rearrangement between two nitrogen atoms (blue). Reduced riboflavin is easily oxidized by O², but an oxygen scavenger system removes it to protect reduced state of riboflavin. (B) Photoconversion efficiency of fixed E. coli expressing EGFP in various surrounding media. The measurements are done as in Figure S1. OS, oxygen scavenger; GC, glucose oxidase/catalase; PCD, protocatechuate-3,4-dioxygenase; RiM, 0.1 mM riboflavin and 0.5 mM methionine. (C) Photoconversion efficiency of fixed E. coli expressing EGFP in the modified RiMOS (riboflavin, methionine and oxygen scavenger). Ri, 0.1 mM riboflavin; Met, methionine; FAD, 0.1 mM flavin adenine mononucleotide. Oxygen scavenger is included in all media. (D) Fluorescence emission spectra of EGFP (excitation with 532 nm). (E) Absorption spectra of EGFP in response to the change in the surrounding environment. PBS was used as blank for EGFP+PBS whereas RiMOS was used as blank for series of measurements using EGFP+RiMOS. Inset shows absorption spectra of a RiMOS solution where the color code is the same as EGFP, and PBS was used as blank (F) Histogram of number of photons emitted from EGFP in both green and red forms measured by single molecule imaging.

Figure 2. A Simulation of PALM Reconstruction in Relation to the Noise.

(A) Points making up a simulated helix with a pitch of 80 nm is confined in a 5,600 nm ×160 nm ×160 nm space, and randomly scattered over 10,000 time frames. This is convolved with a real point-spread function (PSF) to imitate real microscopic diffraction spots. The central cross section was used as a raw PALM image. (B) “Sum” is the summation of all diffraction limited raw images, and “PALM” is the PALM reconstruction. (C) Acquired noise image with SD of 1.67 photons (SNR 7.13). The green box shows where the helix is to be embedded. Bar is 1.0 µm. (D) Images of different noise levels were added to the simulated raw image. Shown here are results for three noise levels corresponding to SNRs of 12.43, 7.13, and 3.29. For each noise level, shown in color are a raw image with a representative peak and the same image denoised, denoised and then deconvolved, and median-filtered and then deconvolved. Shown in grayscale are the PALM reconstructions corresponding to each of the 4 circumstances. The pixel size of the PALM reconstruction images is 1/6 of that of the raw images. (E) One-dimensional localization precision as full width of half maximum (FWHM) of the error distribution (see Fig S2A). Localization precision from the raw and the denoised images is outside of plot area at SNR 3.29 due to the unavailability of enough number of points to calculate FWHM. (F) Point finding efficiency. The graph color code in (E) applies to this plot as well.

Figure 3. Drosophila Metaphase Chromosome Arms Consist of Fine Chromatin Fibers.

(A, B) PALM reconstruction of chromosome arms boxed in (C) free from cytoplasm. (D) PALM reconstruction of chromosome arms boxed in (E) surrounded by the cytoplasm. Red arrowheads point to typical filamentous structures, although this structure is almost everywhere in the chromosome arms. (C, E) Denoised wide-field images of chromosomes. DAPI staining is shown in purple and H2AvD-EGFP is shown in green. Bar shown in (A) corresponds to 0.5 µm in (A, B, D) and 3 µm in (C, E). (F) A schematic drawing of the metaphase chromosome arm based on the interpretation of the PALM reconstruction.