Multiplexed and high-throughput neuronal fluorescence imaging with diffusible probes

Abstract

Synapses contain hundreds of distinct proteins whose heterogeneous expression levels are determinants of synaptic plasticity and signal transmission relevant to a range of diseases. Here, we use diffusible nucleic acid imaging probes to profile neuronal synapses using multiplexed confocal and super-resolution microscopy. Confocal imaging is performed using high-affinity locked nucleic acid imaging probes that stably yet reversibly bind to oligonucleotides conjugated to antibodies and peptides. Super-resolution PAINT imaging of the same targets is performed using low-affinity DNA imaging probes to resolve nanometer-scale synaptic protein organization across nine distinct protein targets. Our approach enables the quantitative analysis of thousands of synapses in neuronal culture to identify putative synaptic sub-types and co-localization patterns from one dozen proteins. Application to characterize synaptic reorganization following neuronal activity blockade reveals coordinated upregulation of the post-synaptic proteins PSD-95, SHANK3 and Homer-1b/c, as well as increased correlation between synaptic markers in the active and synaptic vesicle zones.

Fig. 1

Schematic of the PRISM imaging technique. a Reagents (markers, shown in gray) for detecting subcellular targets include antibodies or peptides that are conjugated with unique oligonucleotide barcodes (docking strands, shown in blue). A barcoded marker is imaged using the complementary fluorophore-conjugated oligonucleotides (imaging probes) that bind to the docking strands on the marker (see (iii) in b, fluorophores shown as red circles). Binding affinity of the imaging probes to the docking strands can be varied by changing the sequence and type of the oligonucleotides, which thereby enables either diffraction-limited (high affinity) or super-resolution microscopy (low affinity). Conjugation of docking strands to markers using site-specific click chemistry enables stoichiometric control of the number of nucleic acids bound to a whole antibody, while SMCC enables conjugation of docking strands to free amine groups on a variety of markers. b The reagent testing and validation phase consists of (i) generation of reference staining patterns of all molecular targets of interest using standard immunofluorescence (IF), (ii) analysis of the specificity and staining quality of markers conjugated with docking strands compared to those in the reference IF, and (iii) co-localization of PRISM-imaged staining patterns using imaging probes (red circles, which correspond to fluorophores conjugated to the probes) with standard IF staining patterns (green circles). c Overview of the main steps in the PRISM imaging workflow. All molecular targets of interest are immunostained at once using docking strand-conjugated markers (e.g., antibodies shown in green, blue, and pink). Nucleic acid imaging probes specific to each marker (e.g., p₁–p₁₀) are applied and imaged sequentially, with each imaging strand washed out after image acquisition at each step. This approach enables imaging a dozen or more distinct molecular targets in the same sample. Images of different markers are drift-corrected and overlaid to generate a pseudo-colored, multiplexed image. For super-resolution PRISM, prior to drift correction, the super-resolved image of each marker is reconstructed from the temporal image stack of binding/unbinding events of the imaging probes to/from the docking strands on the marker

Fig. 2

Blocking off-target nuclear localization of ssDNA-conjugated antibodies. a Neurons were stained with native or ssDNA-conjugated anti-bassoon antibody, anti-synapsin-I antibody, and DAPI. ssDNA-conjugated anti-bassoon antibody exhibited strong off-target nuclear localization (ii, green staining inside the nuclei), compared to the native antibody (i). This nuclear localization was reduced by blocking the fixed sample with non-specific (salmon sperm) DNA prior to immunostaining (iii) or when the anti-bassoon antibody used for staining was conjugated with ssPNA instead of ssDNA (iv). Scale bars: 20 μm. b Cross-correlation analysis of the IF images in a. Pearson correlation coefficient (PCC) of the bassoon channel (green in a) with the synapsin-I channel (red in a) for each image. Differences in PCC indicate changes in antibody staining patterns. Error bars represent 95% confidence intervals

Fig. 3

Confocal LNA-PRISM images of rat hippocampal neuronal synapses. a 13-channel images of 21 days in vitro (DIV) rat hippocampal neuronal culture from a single field of view. Individual channels for each marker are shown followed by a composite image in the bottom-right corner. MAP2 and VGLUT1 were visualized using fluorescently labeled secondary antibodies, nuclei were visualized using DAPI, and all other targets were visualized using ssLNA imaging probes. Synapsin-I was imaged twice, once in the middle and once at the end of the experiment. b Zoom-in view of a single dendrite indicated by the white box in a. Scale bars: a 20 μm; b 2 μm

Fig. 4

Analysis of single-synapse profiles from confocal imaging. a LNA-PRISM images of a conventional excitatory synapse (yellow arrowhead) with co-localization of every synaptic marker measured, and a synapse with only a subset of markers present (white arrowhead). Scale bar: 1 μm. b Network representation of correlations between intensity levels of synaptic proteins within synapses (n = 178,528 synapses from three cell culture batches). The thickness of each edge represents the relative correlation strength between the respective nodes. c t-Distributed Stochastic Neighbor Embedding (t-SNE) plots of n = 10,000 synapses from a single culture batch; each with 20 features (intensity levels and punctae sizes of ten synaptic proteins). Each point in each t-SNE map represents a single synapse with its (x,y) coordinates corresponding to the transformed features that best preserve the distribution of synapses in the original high dimensional feature space. Intensity levels of individual proteins are color-coded in each map. E: cluster of conventional excitatory synapses with the presence of most synaptic markers; I: cluster of possible inhibitory synapses with the absence of most synaptic markers; S: cluster of possible sub-type synapses with the presence of only a subset of synaptic markers. d Hierarchical clustering analysis of synapse profiles. Each column in the heat map represents a profile of a single synapse with 24 synaptic features (rows). I and A denote image intensity level and punctum size, respectively (n = 53,698 synapses from a single culture batch)

Fig. 5

Analysis of synaptic remodeling from confocal imaging. a Representative images of synapses from 21 days in vitro rat hippocampal neurons. White arrowheads indicate synapses. Scale bar: 1 μm. b Bar heights indicate the average synaptic integrated intensities (top) or average synaptic punctae areas (bottom) of untreated and TTX treated neurons relative to the mean of the untreated group. Error bars represent 95% confidence intervals. P-values are computed using Student’s t-test with n = 6 replicates; p < 0.05 (*), p < 0.01 (***). c–e t-SNE plots of individual synapses from untreated (n = 10,000) and TTX treated (n = 10,000) synapses. Mean intensities and areas from each target are used as input. c Kernel density estimate of t-SNE output. White points indicate potential synaptic sub-types. d Individual points color-coded to indicate untreated (white) or TTX treated (black) synapses. e Individual points are color-coded for Homer-1b/c mean intensity and Homer-1b/c area

Fig. 6

Super-resolution DNA-PRISM imaging of primary neuronal cultures. a Widefield and DNA-PRISM images of neuronal microtubules stained using the DNA-conjugated anti-Tuj-1 antibody. b Zoom-in view of the boxed areas in (a) show resolution enhancement of DNA-PRISM images compared with widefield images. The arrowhead indicates distinct microtubule bundles that are not resolved in the widefield image. c Widefield and DNA-PRISM images of filamentous actin stained using DNA-conjugated phalloidin. d Zoom-in views of the boxed areas in c show two actin filaments (left) and the synaptic actin punctae with sub-synaptic structures (right, arrowhead) that are not resolved in widefield images. e Widefield and DNA-PRISM images of pre-synaptic marker synapsin-I (red) and post-synaptic marker PSD-95 (cyan) of the same field of view. f Zoom-in view of single synapses indicated by boxes in e. g Cross-sectional profile of the boxed region in (b) shows a microtubule bundle next to a possible single microtubule with FWHM = 47 nm. h Cross-sectional profile of the boxed region in d shows two actin filaments or small filament bundles that are 80 nm apart. i The average size of synapses defined by synapsin-I and PSD-95 is quantified using the normalized radial cross-correlation function. The decay at the smaller radial shift of the DNA-PRISM curve (red) indicates the smaller synapse size in the DNA-PRISM image due to the improved spatial resolution. Scale bar: a, c, e 10 μm; b, d, f 0.5 μm

Fig. 7

Distributions of synaptic proteins within individual synapses. a DNA-PRISM images of neurons showing the subset channels of synaptic proteins (left), the subset channels of cytoskeletal proteins (middle), and all the channels (right). b Zoom-in view of two individual synapses in a shows the separation of pre-synaptic proteins (synapsin-I and bassoon) and post-synaptic proteins (PSD-95, SHANK3, Homer-1b/c). For each synapse, the nine-target image is shown in the top-left corner, with distinct pairs of synaptic proteins shown in the remaining images. Synapsin-I was imaged twice, once at the beginning and once at the end of the experiment. c Cross-sectional profiles of protein distributions along trans-synaptic axes (white boxes with arrows in b) of the two synapses in b. Red lines indicate the medians of the distributions. Scale bars: 10 μm in full field views; 500 nm in zoom-in views