High-performance probes for light and electron microscopy

Abstract

We describe an engineered family of highly antigenic molecules based on GFP-like fluorescent proteins. These molecules contain numerous copies of peptide epitopes and simultaneously bind IgG antibodies at each location. These 'spaghetti monster' fluorescent proteins (smFPs) distributed well in neurons, notably into small dendrites, spines and axons. smFP immunolabeling localized weakly expressed proteins not well resolved with traditional epitope tags. By varying epitope and scaffold, we generated a diverse family of mutually orthogonal antigens. In cultured neurons and mouse and fly brains, smFP probes allowed robust, orthogonal multicolor visualization of proteins, cell populations and neuropil. smFP variants complement existing tracers and greatly increase the number of simultaneous imaging channels, and they performed well in advanced preparations such as array tomography, super-resolution fluorescence imaging and electron microscopy. In living cells, the probes improved single-molecule image tracking and increased yield for RNA-seq. These probes facilitate new experiments in connectomics, transcriptomics and protein localization.

Figure 1

Probe development and preliminary characterization. (a) Schematic of modular smFP construction. sfGFP shown in cartoon; chromophore either left intact or rendered invisible (grey). Epitope tags are inserted at the N- (Ins 1, blue sphere) and C- (Ins 3, red) termini and into a loop (Ins 2, yellow). (b–c) Fluorescence correlation spectroscopy (FCS) measurements to quantify antibody binding to FLAG epitopes of smFP, based on changes in diffusion with molecular weight. (b) Calibration of solution-phase diffusion time τ_D and diffusion coefficient D versus molecular weight (MW) measured by FCS. Molecular weight markers were: hydrolyzed Alexa488, 534 Da; hydrolyzed Alexa546, 963 Da; epidermal growth factor (EGF)-FITC, 6.5 kDa; EGFP, 32.7 kDa; smFP_FLAG_bright, 42.3 kDa; bovine serum albumin (BSA)-Alexa488, 69 kDa; anti-FLAG M2-FITC, 153 kDa. n = 5 experimental replicates; mean ± SD shown. Data was fit to a power-law: τ_D = (0.072 ± 0.006) ×MW^(0.39±0.02) (± SEM, R² = 0.99). Diffusion time and diffusion coefficient are related by D = w_o²/8τ_D, where w_o is the e⁻² laser-beam radius, here w_o = 430 nm at 940 nm excitation. Also shown are literature values of diffusion coefficients of proteins, and dyes and proteins measured using scanning FCS. c) FCS-determined diffusion times of 10 nM smFP_FLAG_bright (with 10 FLAG copies per protein yielding 100 nM of FLAG epitopes) titrated with monoclonal antibodies against GFP (Invitrogen rabbit monoclonal) or FLAG (Sigma M2), which shows saturated binding above 100 nM antibody; and 10 nM smFP_3xFLAG_bright (30 nM epitopes) with anti-FLAG antibody, which shows saturated binding above 30 nM antibody. n = 5 experimental replicates for titrations; mean ± SD shown. Ladder at right is the calculated number of antibodies bound to smFP (42.3 kDa) for the corresponding τ_D, based on the calibration obtained in c). d) Fly brains showing expression and staining of smFP probes. R59A05-GAL4 crossed with UAS-myr-smFP flies, with myr-sfGFP as a control. All probes, including sfGFP, lack chromophores. Dissected fly brains were stained with anti-tag antibodies or anti-GFP and Alexa488-conjugated secondaries, in small volumes (~10 µl); slight differences in staining are likely due to variation in antibody penetration. Scale bar, 50 µm.

Figure 2

Multi-channel projection labeling with smFP probes and FPs. (a) Injection schematic. Right hemisphere injected with smFP_FLAG and left hemisphere injected with smFP_myc. smFP_FLAG and smFP_myc are detected with tag-specific primary antibodies and corresponding secondary antibodies conjugated with Alexa488 (myc) and Alexa594 (FLAG). (b) Representative (n = 2 replicates) injection site image. (c) Confocal image showing a zoomed view of boxed area in panel b; long-range axonal projections from S1 in both hemispheres crossing the corpus callosum. (d) Injection schematic for the 4-color tracing. Left hemisphere of Rbp4_KL100_Cre mouse injected (all AAV-FLEX-CAG) with two fluorescent proteins (tdTomato and GFP) as well as two smFP constructs (smFP_myc and Ruby2_FLAG). Injections targeted four topographic areas: vibrissal sensory (vS1) and vibrissal motor cortex (vM1) as well as limb motor (fM1) and sensory areas (llS1) as indicated. (e–n) Schematics and images of 80 µm coronal sections. (e–f) Injection sites. Fluorescent neuronal somata in L5 at injection sites in fM1/vM1 (h) and vS1/llS1 (i) are clearly visible, as are long-range axonal projections. Schematics of long-range targets of L5 neurons are provided (g, j, m) alongside images of coronal sections from these planes. Red boxes indicate enlarged areas. (h, i) show cortico-cortical and cortico-thalamic axons in ectorhinal (Ect) and perirhinal (PRh) cortex as well as thalamus (POm). These descending axons continue in adjacent tracts to the cortical peduncle (cp; h–l) and target midbrain regions including substantia nigra reticulata (SNR) and superior colliculus (SC; k, l), as well as the spinal trigeminal nucleus (Sp5; m, n), >5 mm from the injection site. Intermingling of axons of all four colors in the pyramidal tract (py) results in the bright white color in n. Drawing are adapted with permission from Paxinos and Franklin, Mouse Brain in Stereotaxic Coordinates, 2nd Ed. (2001).

Figure 3

Improved labeling of cells and single molecule tracking efficiency of proteins in fixed and live preparations. Comparison of smFP_FLAG (a, b) and EGFP (c, d) labeling of CA1 hippocampal pyramidal cells. (a, c) Fields of view showing sparsely labeled cells, from apical dendrites through soma to distal dendrites. (b, d) Zoom of distal dendrite in boxed areas shown in a and c. All images were taken under identical confocal settings. Blue shows Hoechst nuclear counterstain. (e–g) Images of CA3 “thorns” in P15 mouse brain slices, expressing either smFP_FLAG (e) or EGFP (f) or filled with Lucifer Yellow (LY) (g). Samples were amplified with anti-FLAG, anti-GFP or anti-LY primary antibodies and secondary antibodies conjugated to Alexa488. (h–k) Single dendrites of cultured rat hippocampal neurons co-electroporated with smFP_myc and either N-cadherin-smFP_HA (h, j) or N-cadherin-HA (i, k). The smFP_myc fills cells cytoplasmically and serves as a control for normalizing expression levels. Blue: anti-myc primary + secondary-Alexa488, pseudocolored blue. Red: anti-HA primary + secondary-Alexa555. h, i: combined imaging channels; j, k: N-cadherin channel alone. (l–o) HeLa cells with tracked labeled histone molecules. All images were acquired and displayed with identical parameters. (l) H2B-smFP_FLAG transfected cell with Alexa488-labeled anti-FLAG antibody Fab fragments. (m) H2B-EGFP transfected cell with Alexa488-labeled anti-GFP antibody. (n) H2B-Halo tag transfected cell with Alexa488 Halo substrate added. (o) Background labeling of anti-FLAG antibody. Alexa488-labeled anti-FLAG antibody added to untransfected cells. (p) Histogram of the intensity of single molecule tracks from EGFP (green), smFP_FLAG (magenta) and Halo tag (black). Arrowheads point to the smFP_FLAG histogram peaks (intensities ~7k; 14k; 21k; 28k; 35k; 42k; 49k) that give an estimate of the number of bound Fab molecules per tag. The smFP_FLAG histogram is fit by $f(I)={\sum }_{n=1}^{10}{C}_n^{10}{p}^n{(1-p)}^{10-n}\frac{I^{2n-1}{a}^{2n}}{(2n-1)!}{e}^{-aI}$ with a determined from the fit to the Halo tag histogram (n = 1), and with p = 0.2, consistent with stochastic antibody binding with an average Fab occupancy of 20% (dashed curve). The three dotted curves depict the first three components of f(I). Movies corresponding to l–o are provided in Supp. Movie 1.

Figure 4

Utility of smFP_FLAG for array tomography. (a) Left: large-scale AT z-projections showing smFP_FLAG labeling of somatostatin-Cre interneurons with cell bodies in CA1 hippocampus stratum oriens (SO) and axonal projections in stratum lacunosum moleculare (SLM). Center: EGFP+ pyramidal cell signal from the same volume. Right: merged projection. (b) A series of 3 consecutive ultra-thin sections (100 nm) showing an smFP_FLAG-labeled varicosity (magenta) forming a putative synapsin (gray) immunopositive synapse onto a distal pyramidal cell dendrite (green) in SLM. Note the labeling consistency across serial sections for each channel. Bottom: zoomed images showing each channel from section two (white box), with arrows showing pixels overlaying between the synapsin and smFP_FLAG channels.

Figure 5

Utility of smFPs for STORM and immuno EM. (a–d) Multi-color STORM imaging of smFPs in mouse brain slice. (a) STORM image of IHC-labeled neurites expressing tdTomato and smFP_myc constructs from an ultracryosection of mouse cortex. (b) smFP_myc and tdTomato yield comparable STORM images of individual dendritic spines on this layer V pyramidal neuron (bottom panels correspond to dashed boxed region in a. The diffraction-limited conventional image counterparts are shown on the top for comparison. Scale bars: a, 5 µm; b, 500 nm. (c) Schematic of the double immunogold labeling experiments in mouse brain tissue. Confocal image shows a representative vibratome section expressing green smFP_FLAG_bright fluorescence and immunolabeled with anti-myc (red) to verify expression of both constructs (scale bar 20 µm). smFPs were double-labeled with 6 nm (myc) and 12 nm (FLAG) colloidal gold particles followed by silver enhancement. HPF-FS, high-pressure freezing followed by freeze substitution. (d–e) smFP_FLAG immunogold labeling (yellow pseudocolor) and the resulting serial reconstruction (41 serial sections) of dendrites from a layer 2/3 cortical neuron. Scale bar 1 µm. (f–g) Double immuno EM labeling and the resulting serial reconstruction of dendrites from two layer 2/3 cortical neurons (30 serial sections). smFP_FLAG pseudocolored yellow, smFP_myc pseudocolored red. Note high density of stain in all images, even in spines, with minimal background labeling. Scale bar 1 µm. (h–i) Silver enhanced anti-HA immunogold labeling in HeLa cells: (h) 1% OsO₄, (i) No secondary fix. Notice the improved structure preservation with 1% OsO₄, with no observable difference in label density. Movies showing the 3D reconstruction of the data in d–g are provided in Supp. Movies 2–5.