Supernova: A Versatile Vector System for Single-Cell Labeling and Gene Function Studies in vivo

Abstract

Here we describe "Supernova" series of vector systems that enable single-cell labeling and labeled cell-specific gene manipulation, when introduced by in utero electroporation (IUE) or adeno-associated virus (AAV)-mediated gene delivery. In Supernova, sparse labeling relies on low TRE leakage. In a small population of cells with over-threshold leakage, initial tTA-independent weak expression is enhanced by tTA/TRE-positive feedback along with a site-specific recombination system (e.g., Cre/loxP, Flpe/FRT). Sparse and bright labeling by Supernova with little background enables the visualization of the morphological details of individual neurons in densely packed brain areas such as the cortex and hippocampus, both during development and in adulthood. Sparseness levels are adjustable. Labeled cell-specific gene knockout was accomplished by introducing Cre/loxP-based Supernova vectors into floxed mice. Furthermore, by combining with RNAi, TALEN, and CRISPR/Cas9 technologies, IUE-based Supernova achieved labeled cell-specific gene knockdown and editing/knockout without requiring genetically altered mice. Thus, Supernova system is highly extensible and widely applicable for single-cell analyses in complex organs, such as the mammalian brain.

Figure 1. IUE-based Supernova enables sparse and bright cell labeling with little background.

(a) Schematics showing the elementary components of IUE-based Supernova: TRE-SSR-WPRE-pA (TRE-SSR) and CAG-RT-stop-RT-XFP-ires-tTA-WPRE-pA (CAG-RT-stop-RT-XFP-tTA). TRE: tetracycline response element; tTA: tetracycline transactivator; SSR: site specific recombinase (e.g. Cre, Flpe); RT: recombination target site (e.g. loxP, FRT); XFP: fluorescent proteins (e.g. GFP, RFP); WPRE: WHP post-trascriptional response element; ires: internal ribosome entry site. (b) Schematics showing how Supernova works: (1) Initially, only in a sparse population among many cells that are transfected with both vectors, the leakage of TRE drives above-threshold but weak SSR expression. (2) This low level of SSR excises the RT-stop-RT cassette in a few copies of CAG-RT-stop-RT-XFP-tTA vector, initiating the transcription of XFP and tTA, albeit weakly. (3) Through binding with TRE, tTA facilitates expression of SSR. (4) Then RT-stop-RT cassette is excised from many copies of CAG-RT-stop-RT-XFP-tTA vector, and expression of XFP and tTA is increased. This positive loop of tTA/TRE enhancement (See Supplementary Fig. 3) leads to extremely high levels of expression of both SSR and XFP, only in a small population of transfected cells. (c–g) Supernova labeling is sparse and bright enough to visualize the detailed morphology of single neurons including dendritic spines (f) and axonal branches and boutons (g). An Flpe-based Supernova GFP (Flpe-SnGFP) vector set (TRE-Flpe and CAG-FRT-stop-FRT-GFP-tTA) was introduced into layer 2/3 (L2/3) cortical neurons by in utero electroporation (IUE). The CAG-RFP vector was co-electroporated to mark the transfected neurons. Higher-power images of the rectangles in c were shown in (d). The square in (d) were further magnified in (e). (f,g) Show higher-magnification images of the rectangles in (e). (h) Two-photon in vivo imaging of L4 cortical neurons labeled by Flpe-based Supernova RFP (Flpe-SnRFP) in P5 mouse. The traces of imaged cortical neurons were shown in right panel. Black lines indicate the dendrites of labeled neurons. The axons of these neurons are represented by red and blue lines, separately. Scale bars, 250 μm (c); 100 μm (d); 50 μm (e,h); 4 μm (f); 10 μm (g).

Figure 2. The sparseness of Supernova labeling is stable and adjustable.

(a,b) The sparseness and brightness of Supernova labeling were constant from early postnatal stages to adulthood. Images of L2/3 cortical neurons labeled by Flpe-SnGFP, in which concentration of the TRE-Flpe vector was 5 ng/μl (standard concentration), were shown (a). CAG-RFP was co-electroporated to label all the transfected cells. Coronal sections were made from P8, P22, 2 month-old (2M), 4M and 8M brains. The ratio of Flpe-SnGFP-labeled cell number to transfected cell number at each age was shown under the image in (a) and as diagram in (b) [mean ± SEM; n = 5 mice at P8 (30/2162 cells), P22 (8/654 cells), 2M (8/562 cells), n = 3 mice at 4M (9/529 cells), n = 2 mice at 8M (7/555 cells)]. Scale bars, 100 μm. (c,d) The degree of sparseness of Supernova-labeling is adjustable. Concentrations of TRE-Flpe vector used for IUE were 5 ng/μl, 50 ng/μl and 500 ng/μl [n = 5 mice for each group; 30/2162 cells (5 ng/μl), 890/1880 cells (50 ng/μl), 2507/2539 cells (500 ng/μl)]. L2/3 cortical neurons were labeled by Flpe-SnGFP together with CAG-RFP. Data obtained using 5 ng/μl are common in panels (a,c). Scale bars, 100 μm.

Figure 3. Simultaneous visualization of multiple proteins in a single cell using the Supernova.

(a) Using the Supernova, RFP and GFP were expressed in sparsely labeled neurons with high co-expression efficiency (RFP⁺/GFP⁺ = 51/53 cells, GFP⁺/RFP⁺ = 51/51 cells, n = 5 mice). Flpe-based Supernova vector sets (TRE-Flpe, CAG-FSF-RFP-tTA and CAG-FSF-GFP-tTA) were introduced by IUE. (b) RFP fused with nuclear localization signal (nlsRFP) and GFP was co-expressed in a small population of cortical neurons by introducing Supernova vectors (TRE-Cre, CAG-LSL-GFP-tTA and CAG-LSL-nlsRFP-tTA). (c) Simultaneous visualization of RFP and PSD95-GFP in the same single neuron. Upper panel shows low-power images of a tangential section through the somatosensory cortex L4 transfected with SnRFP and Supernova PSD95-GFP (SnPSD95-GFP) vectors (TRE-Cre, CAG-LSL-RFP-tTA and CAG-LSL-PSD95-GFP-tTA). Bottom shows higher-magnification images of the rectangles in the upper panel. Scale bars, 100 μm (a,b); 25 μm (c).

Figure 4. Labeled cell-specific gene knockout via Cre-based Supernova in floxed mice.

(a) Schematic for Supernova-mediated single-cell knockout (KO) of endogenous gene of interest (GOI) flanked by two loxP sites. (b) α2-Chimaerin protein is expressed ubiquitously in the hippocampal CA1, while it is undetected specifically in SnRFP-labeled neurons (arrows), indicating that Cre-based Supernova-mediated gene knockout is highly specific to the labeled cells. Cre-SnRFP vectors were introduced into α2-chimaerin (α2-Chn)^flox/flox mouse CA1 by IUE. α2-chimaerin immunohistochemistry and DAPI-staining were performed. Upper panels show the hippocampus of α2-Chn^flox/flox mouse. A set of enlarged example images is shown in the bottom. Note that because extremely high intensity signal of Supernova labeling hinders detection of α2-chimaerin signal, we partially photobleached SnRFP signal in this experiment. (c,d) Quantification of Supernova-dependent α2-Chn knockout efficiency and specificity. (c) α2-Chimaerin expression was detected in almost all of SnRFP-negative CA1 hippocampal cells (97.7% ± 0.1%, n = 3 mice; 785 cells/804 cells, 464 cells/474 cells, 1011 cells/1036 cells), while only in 5.9% ± 3.0% (n = 3 mice; 3 cells/31 cells, 0 cell/18 cells, 3 cells/37 cells) of SnRFP-positive CA1 neurons, indicating labeled cell-specific knockout. All values represent as mean ± SEM; (**): 0.001 < P < 0.01; Welch’s t-test. (d) Intensities of α2-chimaerin signal in RFP⁺ cells (α2-Chn KO cells) and RFP^- cells (control cells) that surround RFP⁺ cells were measured (n = 14 cells, 3 mice for each group). (***) P < 0.001, Welch’s t-test. Scale bars: 500 μm (b, upper); 100 μm (b, bottom).

Figure 5. Supernova-mediated RNAi achieved sparsely labeled cell-specific gene knockdown in vivo.

(a) Schematic representation of the Supernova vector set that is for single-cell knockdown of gene of interest (GOI). shRNA against GOI was expressed by CAG-LSL-mir30, which is a Cre-dependent shRNA expression vector. (b) CAG-LSL-mir30:GFP-RNAi-Scramble (control: left panels) or CAG-LSL-mir30:GFP RNAi (right panels) was co-electroporated with Supernova nuclear localization signal RFP (Sn-nlsRFP) into CAG-loxP-CAT-loxP-GFP reporter mouse cortex by IUE at E14.5. Sections were stained with an anti-GFP antibody. Scale bar, 50 μm. (c) GFP signals in GFP-RNAi-expressing neurons were significantly lower than those in cells expressing GFP-RNAi-Scramble-expressing neurons (control). Box-and-Whisker plot: the bottom and top of the box are the first and third quartiles; the horizontal line inside the box represents the median; the mean is plotted as a cross. The ends of the whiskers represent the lowest datum within 1.5 interquartile range (IQR) of the lower quartile, and the highest datum within 1.5 IQR of the upper quartile. Data not included between the whiskers were plotted as pluses. (***) < P < 0.001, Brunner-Munzel test. (d) CAG-LSL-mir30:LacZ-RNAi-Scramble (control: left panels) or CAG-LSL-mir30:LacZ-RNAi (right panels) was co-transfected with SnAmCyan into the Rosa-loxP-stop-loxP-nlsLacZ (RNZ) reporter mouse cortex by IUE at E14.5. Coronal sections were stained with an anti-β-gal antibody to detect LacZ expression. Scale bar, 50 μm. (e) The LacZ expression level was significantly reduced in SnAmCyan-labeled neurons co-expressed with LacZ-RNAi, compared to that co-expressed with LacZ-RNAi-Scramble. Box-and-Whisker plot is similar as (c). (***) p < 0.001, Brunner-Munzel test.

Figure 6. Labeled cell-specific gene knockout in wild-type mouse brain by Supernova-mediated TALEN.

(a) Schematic representation of a Supernova-mediated TALEN vector set. (b) Hippocampal CA1 was introduced with the Supernova α2-Chn TALEN vector set. Note that we attenuated SnRFP signal for clear detection of α2-chimaerin signal. Two sets of example images were shown. (c) Most RFP-negative neurons (control cells) express high level of α2-chimaerin protein (α2-chimaerin^high cell, e.g. Cell 3). While, in majority of RFP-positive neurons [cells expressing Supernova α2-Chn TALEN (Snα2-Chn TALEN)], the α2-chimaerin protein expression is undetectable (α2-chimaerin^negative cell, e.g. Cell 1). Only a small portion of RFP-positive neurons shows α2-chimaerin protein expression, but very weakly (α2-chimaerin^low cell, e.g. Cell 2). Control: 1426 cells; Snα2-Chn TALEN: 53 cells. See Experimental Procedures for details. (d) Top: wild-type sequence of α2-Chn locus. Bottom: mutation patterns detected in Supernova-labeled cells sorted by FACS; red bases: inserted bases; black dashes: deleted bases. Locations of left and right TALENs (thick blue lines) targeting mouse α2-Chn gene are shown. Blue dashed lines indicate spacer region (17 bp) between two TALEN target sites. Scale bar: 20 μm (b).

Figure 7. Labeled cell-specific gene knockout in wild-type mouse brain by Supernova-mediated CRISPR/Cas9.

(a) Schematic for Supernova-mediated CRISPR/Cas9 vector set. (b,c) Hippocampal CA1 region of wild-type mice electroporated with Supernova-CRISPR/Cas9 control (without targeting sequence in sgRNA) (b) and Supernova-CRISPR/Cas9 targeting mouse Creb1 (SnCRISPR-Creb1) (c). CREB, Creb1 gene product, shows strong and ubiquitous expression in the GFP-negative cells in hippocampal CA1 region (both (b,c). CREB expression was also detected in all GFP-positive cells that expressed Supernova-CRISPR control (111/111 cells, 3 mice; hollow arrows in (b). In contrast, expression of CREB was specifically lacking in almost all GFP-positive cells expressing SnCRISPR-Creb1 (81/83cells, 3 mice; arrows in (c), with a very few exceptions (2/83 cells; hollow arrows, lower panels in (c). (d) Percentage of numbers of CREB-positive and CREB-negative cells in GFP-positive cells (SnCRISPR-control- or SnCRISPR-Creb1-expressing cells). (e) Representative mutations found in the Creb1 locus in GFP-positive cells that were collected from the SnCRISPR-Creb1-transfected cortex by FACS at P1. The sgRNA target location is indicated by thick blue lines. The PAM sequence is marked in blue shades. Black dashes: deleted bases. The number of clones showing corresponding mutation pattern was presented in the right. Scale bars, 50 μm (b,c).

Figure 8. AAV-based Supernova enables sparse and bright neuronal labeling and labeled cell-specific gene knockout in vivo.

(a) Schematic representation of the adeno-associated virus (AAV)-based Supernova vector set, which consists of AAV-TRE-Cre-WPRE (AAV-TRE-Cre) and AAV-EF1α-DIO-tTA-P2A-RFP-WPRE (AAV-EF1α-DIO-tTA-RFP) vectors. DIO: Double-floxed Inverted Open reading frame. (b–g) AAV-based Supernova RFP (AAV-SnRFP) labeled hippocampal CA1 pyramidal neurons (b–d) and L5 pyramidal neurons in the cortex (e–g) so sparsely and brightly that detailed morphologies such as dendritic spines and axonal boutons of labeled neurons were clearly visible. AAV injection was performed at P10-P13, and brains were fixed at 30 days post-infection (DPI). AAV-EF1α-GFP-WPRE (AAV-EF1α-GFP) was co-injected as control. (c) Higher-magnification images of the squares in (b). The rectangle in c was further enlarged in (d). (f,g) Enlarged images of squares in (e). Scale bars, 500 μm (b), 100 μm (c,e), 10 μm (d), 25 μm (f,g). (h) Cre-mediated genomic DNA recombination detected using RNZ reporter mice was specific to AAV-SnRFP-labeled neurons in the hippocampus (upper panels) and cortex (bottom panels). AAV-SnRFP was injected into the hippocampus or cortex of RNZ mice at P10, and brains were fixed at 40DPI. Coronal sections were prepared and stained with an anti-β-gal antibody, which detects LacZ expression, and DAPI. Scale bars, 50 μm. (i) α2-Chn was disrupted specifically in an AAV-SnRFP-labeled neuron in the hippocampal CA1. AAV-SnRFP was injected into the hippocampus of α2-Chn^flox/flox mice at P2, and brains were dissected and sectioned at P18. Immunohistochemistry was performed to detect α2-chimaerin expressing cells. α2-chimaerin was expressed in most cells (DAPI) but not in an AAV-SnRFP-labeled cell (arrow) in the hippocampal CA1. Scale bar, 50 μm. Note that we partially photobleached AAV-SnRFP signal in these experiments (h,i) to avoid high intensity signal of Supernova labeling overwhelms α2-chimaerin and LacZ signals.