Avian sarcoma leukosis virus receptor-envelope system for simultaneous dissection of multiple neural circuits in mammalian brain

Significance

Genetic dissection of multiple neural pathways remains challenging because of the limited number of genetic methods that can be used simultaneously. To overcome this limitation, we used modified avian sarcoma and leukosis virus envelopes and receptors to develop highly orthogonal genetic tools that can achieve expression of different genes in different target cells. From in vitro and in vivo screens, we identified tools that can specifically transfer genes of interest into mammalian neurons via engineered receptors, with minimal unintended interactions. Using this approach, we achieved pathway-specific, differential fluorescent labeling of three thalamic neuronal populations, each projecting into different cortical regions. Thus, our approach provides independent, simultaneous, and specific genetic tools for manipulating intermingled neural pathways in vivo.

Abstract

Pathway-specific gene delivery is requisite for understanding complex neuronal systems in which neurons that project to different target regions are locally intermingled. However, conventional genetic tools cannot achieve simultaneous, independent gene delivery into multiple target cells with high efficiency and low cross-reactivity. In this study, we systematically screened all receptor–envelope pairs resulting from the combination of four avian sarcoma leukosis virus (ASLV) envelopes (EnvA, EnvB, EnvC, and EnvE) and five engineered avian-derived receptors (TVA950, TVB^S3, TVC, TVB^T, and DR-46TVB) in vitro. Four of the 20 pairs exhibited both high infection rates (TVA–EnvA, 99.6%; TVB^S3–EnvB, 97.7%; TVC–EnvC, 98.2%; and DR-46TVB–EnvE, 98.8%) and low cross-reactivity (<2.5%). Next, we tested these four receptor–envelope pairs in vivo in a pathway-specific gene-transfer method. Neurons projecting into a limited somatosensory area were labeled with each receptor by retrograde gene transfer. Three of the four pairs exhibited selective transduction into thalamocortical neurons expressing the paired receptor (>98%), with no observed cross-reaction. Finally, by expressing three receptor types in a single animal, we achieved pathway-specific, differential fluorescent labeling of three thalamic neuronal populations, each projecting into different somatosensory areas. Thus, we identified three orthogonal pairs from the list of ASLV subgroups and established a new vector system that provides a simultaneous, independent, and highly specific genetic tool for transferring genes into multiple target cells in vivo. Our approach is broadly applicable to pathway-specific labeling and functional analysis of diverse neuronal systems.

In the mammalian brain, neurons projecting to different target regions are locally intermingled, and even adjacent neurons can have different connectivities and functions (1–5). To disentangle such complex networks, anatomical mapping and individual manipulation of each neural pathway are both critical. To date, conventional genetic tools, such as site-specific recombinase (Cre/loxP) (6), prokaryotic repressor (Tet-On/Off system) (7), and a viral receptor–envelope pair (TVA–EnvA) (8) have been used to genetically control specific neural pathways in the mammalian brain, usually in combination with retrograde viral infection (9).

Despite their usefulness, these technologies can simultaneously manipulate no more than two neural pathways—a number that is far from sufficient for a dissection of natural brain circuitry. Therefore, it is necessary to increase the number of available tools. One challenge to doing so is that all such tools need to be orthogonal; in other words, the biological reactions on which they are based must not cross-react.

A number of biochemical mechanisms are potentially applicable to orthogonal gene expression systems. One of the most promising candidates is the use of specific combinations of virus envelope (Env) proteins and corresponding receptors. Because viral vectors pseudotyped with Env proteins exclusively infect cells that express compatible receptors, exogenous expression of receptors in target cells provides a specific guide for viral entry. However, some viruses can use more than one molecular species as receptors, and these receptors provide a variety of functions essential for viral entry. In simple situations, receptors bind to virus envelopes and initiate endocytic uptake of viruses; alternatively, receptors affect cellular signaling pathways that facilitate virus entry, or they directly activate fusion/penetration processes by inducing conformational changes in Env proteins (10).

Among the enveloped viruses, we focused on avian sarcoma leukosis virus subgroups (ASLVs) for three reasons. First, ASLV’s natural host range is restricted to birds. Therefore, it is likely that ASLV-pseudotyped viral vectors will exhibit low, nonspecific infectivity toward mammalian neurons via endogenous receptors. Second, ASLVs are reported to require single molecular species as receptors, and this simple mode of infection is suitable for adaptation as a conditional gene-delivery system. Third, many different Env and receptor proteins are available from among six distinct ASLV subgroups (A, B, C, D, E, and J) and 10 different receptors (11), and there is evidence that some ASLV Env proteins exhibit specific binding to disparate receptor sequences. For example, chicken TVA protein belongs to the family of low-density lipoprotein receptors and determines susceptibility to ASLV-A (12). The tumor necrosis factor receptor-related proteins TVB^S1 and TVB^S3 confer susceptibility to ASLV-B/-D/-E, and ASLV-B/-D, respectively (13). ASLV-C uses the TVC protein of the butyrophilin family, which contains two Ig-like domains (14). TVB^T, a turkey homolog of TVB, is an ASLV-E–specific receptor (15). Finally, ASLV-J uses the chicken multi-membrane-spanning cell-surface protein Na⁺/H⁺ exchanger type 1 (chNHE1) as a receptor (16).

However, no study to date has systematically and quantitatively determined which types of ASLVs can infect with high efficiency mammalian cells expressing a single receptor, and no study has shown which combination of envelopes and receptors of ASLV specifically interact with each other. In this study, we first performed in vitro and in vivo screens to identify orthogonal receptor–envelope pairs. We then conducted a proof-of-concept study in vivo to demonstrate that these orthogonal pairs can achieve pathway-specific differential fluorescent labeling of multiple neuronal populations, each projecting to different cortical regions. Our findings expand the repertoire of genetic tools that can be used to dissect and manipulate the complex neural networks created by intermingled neurons projecting to different target regions.

Results

Putative Orthogonal ASLV Receptor–Envelope Pairs.

The various ASLV subgroups exhibit different infectious properties in chicken cell lines (CEFs and DF-1 cells) (12–17). Based on these data, we estimated the feasibility of six envelopes and seven receptors as genetic tools, and eliminated several (TVB^S1, ASLV-D envelope, ASLV-J envelope, and chNHE1) for the following reasons: TVB^S1 is a nonspecific cellular receptor for ASLV-B, -D, and -E (13, 17); ASLV-D can infect a variety of mammalian cell lines in the absence of exogenous receptors (18–20); and chNHE1, the receptor for ASLV-J (16), may alter neuronal membrane potential. Additionally, we selected TVA950, a transmembrane form, from two splice-variant forms of TVA, even though the other form of TVA [TVA800, a chicken glycophosphatidylinositol (GPI)-anchored form] has been used in previous studies (8, 21, 22). We chose the transmembrane form because GPI-anchored proteins contain a signal peptide in the C terminus that is cleaved off and replaced by the GPI-anchor, precluding the use of C-terminal epitope tags. This process narrowed down the set of candidates to four receptors (TVA950, TVB^S3, TVC, and TVB^T) and four envelopes (EnvA, EnvB, EnvC, and EnvE) (Fig. 1A).

Fig. 1.

Selection and construction of ASLV-pseudotyped lentiviral vectors and avian-derived receptors. (A) Schematic illustration of receptors expressed on the cell surface and their specificity for ASLV envelope-pseudotyped lentiviral entry. The ASLV receptors and envelopes used in the following experiments were underlined. (B) Schematic illustration of the altered ASLV receptors and envelopes constructs used in this study. Numbers indicate the position of amino acid residues in mature proteins; signal peptide residues have positive numbers. SU, surface envelope subunit; TM, transmembrane envelope subunit; VSV-G, vesicular stomatitis virus G protein; SP, signal peptide; EC, extracellular domain; MS, transmembrane spanning domain; IC, intracellular domain; HA, epitope tag from influenza virus HA protein; His, histidine residue tag; Ser–Glyx4 linker, 5-amino acid peptide normally used as the linker; DR5, DEATH RECEPTOR 5.

Measurement of Orthogonality Between the Receptor–Envelope Pairs in Vitro.

To examine the orthogonality of the receptor–envelope pairs, we tested the various combinations to determine which ones afforded specific viral entry and transgene expression in human embryonic kidney (HEK) 293T cells. To mitigate the unwanted side effects associated with the expression of an exogenous transmembrane protein in mammalian cells, we deleted a part of the original intracellular domains from each receptor (Fig. 1B). In TVA950, the original intracellular domain, except for four amino acid residues, was deleted (12). In TVB^S3, TVB^T, and DR-46TVB, the intracellular domain, except for 107 amino acid residues, was deleted (13, 17). In TVC, the intracellular domain remained. Epitope tags (HA, c-Myc, V5, and 3× FLAG) were C-terminally fused to the receptors to allow detection by immunohistochemistry. To increase protein half-life according to the mammalian N-end rule (23), the N termini of the receptors were also modified, such that valine was the second amino acid.

We then developed two types of recombinant lentiviral vectors: FuGB2-pseudotyped bicistronic vectors that coexpressed the engineered receptor and Aequorea coerulescens green fluorescent protein (AcGFP1) and ASLV Env-pseudotyped vectors that expressed mCherry (Fig. 2A). The FuGB2-pseudotyped lentiviral vector can transduce neurons by retrograde axonal transport in vivo (24). Some dividing cells can be also transduced in vitro (25). In the present study, this vector was used to express ASLV receptors in both in vitro and in vivo experiments. The cytomegalovirus (CMV) promoter was selected because it drives a high level of protein expression in the mammalian brain (26) and has been used previously for TVA expression (8, 21, 27). We tested all possible receptor–envelope combinations by serially infecting receptor-expressing and Env-pseudotyped vectors into HEK 293T cells. mCherry expression was observed in only five pairs: TVA950–EnvA, TVB^S3–EnvB, TVC–EnvC, TVB^T–EnvB, and TVB^T–EnvE (Fig. 2 B–E). Although the results obtained with TVA950–EnvA, TVB^S3–EnvB, and TVC–EnvC were consistent with those obtained in DF-1 cells, the infection of EnvB vector into TVB^T-expressing cells observed in our assays was inconsistent with previous reports.

Fig. 2.

Characterization of ASLV receptor–envelope pairs in vitro. (A) Schematic representation of the lentiviral vector constructs used in vitro and the experimental outline of serial infections of bicistronic vectors and Env vectors in HEK 293T cells. cPPT, central polypurine tract; CTS, central termination sequence; LTR, long terminal repeat; RRE, Rev responsive element; Ψ, packaging signal; CMV, CMV promoter; P2A, picornaviral 2A peptide; WPRE, Woodchuck hepatitis virus posttranscriptional regulatory element. (B–G) Fluorescence images of HEK 293T cells expressing TVA950 (B), TVB^S3 (C), TVC (D), TVB^T (E), and DR-46TVB (F) after transfection with one of the Env vectors. Receptor proteins were coexpressed with AcGFP1. (G) A vector expressing only AcGFP1 was used as a negative control. (Scale bars: 50 μm.)

We next sought to rationally design an alternative EnvE-specific receptor for TVB^T. The chimeric protein DR-46TVB was engineered by exchanging the first cysteine-rich domain of chicken TVB^S1, which is critical for ASLV-B infection (28), with those of the human TVB homolog DEATH RECEPTOR 5. HEK 293T cells expressing this chimeric receptor expressed mCherry in combination with EnvE vector, but not EnvB vector (Fig. 2F).

These results were quantitatively confirmed by flow cytometry. A progressive shift to the right was observed only in the histograms for the following six pairs: TVA950–EnvA, TVB^S3–EnvB, TVC–EnvC, TVB^T–EnvB, TVB^T–EnvE, and DR–46TVB–EnvE (Fig. 3A). Furthermore, the infection rates (the ratio of AcGFP1/mCherry double-positive cells to all AcGFP1-positive cells under each condition) for TVA950–EnvA, TVB^S3–EnvB, TVC–EnvC, and DR-46TVB–EnvE were significantly higher than those of the other pairs [one-way analysis of variance (ANOVA) and Scheffé’s F-test, P < 0.001] except for TVB^T–EnvB and TVB^T–EnvE (Fig. 3B and Table S1). Together, these data demonstrate that these four pairs (TVA950–EnvA, TVB^S3–EnvB, TVC–EnvC, and DR-46TVB–EnvE) are orthogonal in HEK 293T cells.

Fig. 3.

Quantitative comparison of in vitro infection rates of all ASLV receptor–envelope pairs. (A) Representative histograms of mCherry fluorescence intensity for every possible receptor–envelope pair expressed in HEK 293T cells. Receptor-expressing cells infected by EnvA-, EnvB-, EnvC-, or EnvE-pseudotyped vectors (blue, green, yellow, and red histograms) are compared with receptor-expressing cells infected with no Env vectors (purple histograms). In the control condition, HEK 293T cells were infected with AcGFP1-expressing vector instead of the bicistronic vector. The vertical axis represents the total number of counted cells, and the horizontal axis represents Logicle scaling (29). (B) Histograms showing the infection rates of Env vectors into HEK 293T cells positive for each receptor. Cells expressing AcGFP1 were considered receptor-positive. Thus, the infection rate is estimated by calculating the ratio of mCherry- and AcGFP1-positive cells to all AcGFP1-positive cells in flow-cytometry analysis. Three independent experiments were conducted, and the data were analyzed by using one-way ANOVA followed by Scheffé’s F-test. *P < 0.001. The average mean values of three experiments are shown, with the SD of the data indicated by error bars.

Orthogonality of the TVA950–EnvA Pair in Vivo.

Having demonstrated that the four aforementioned orthogonal pairs were capable of transferring genes into cultured mammalian cells, we next examined the gene-transfer capability and the orthogonality of our receptors and envelopes in vivo using the thalamocortical system of rats. FuGB2-pseudotyped lentiviral vector shows a high efficiency of gene transfer via retrograde axonal transport. In these in vivo experiments, the receptor genes were transferred into both thalamocortical and cortico-cortical neurons projecting to the injection site (the primary somatosensory cortex) (Fig. S1). By contrast, fluorescent protein genes were expressed only in the thalamocortical neurons by the injection of Env vectors into the thalamus (Fig. S2). In a pilot study, we injected FuGB2–TVC vector along with a mixture of EnvA/EnvB/EnvC/EnvE vectors into 12 rats. We did not observe any sign of infection by the EnvC vector in these rats in vivo. Hence, the TVC–EnvC pair was not used in the following experiment.

We first tested the specificity of the TVA950–EnvA pair (Fig. 4A). In these experiments, the retrograde vector expressing TVA950 (FuGB2–TVA950) was injected into the primary somatosensory cortex (S1) (Fig. 4B), and a mixture of EnvA/EnvB/EnvE vectors that expressed different fluorescent proteins (XFPs) was injected into the thalamus [EnvA vector expressing blue fluorescent protein (BFP), EnvB vector expressing enhanced green fluorescent protein (EGFP), and EnvE vector expressing tdTomato]. The Env vectors were injected 3 wk after the injection of the FuGB2-pseudotyped vector. This timing was necessary to allow time for TVA950 expression, which mediates the EnvA vector infection. In sections in the immediate vicinity of either the FuGB2–TVA950 or Env vectors’ injection site, many BFP-positive axons in S1 or cells in the thalamus were observed, respectively (Fig. 4C). Near the FuGB2–TVA950 vector injection site, dense projections of BFP-positive axons were observed in layers I, IV, and VI (Fig. 4D), consistent with the projection pattern of thalamocortical neurons (30–34). By contrast, around the injection sites of the Env vectors, BFP-positive cells were detected in the ventral posteromedial thalamic nucleus (VPM) and the posteromedian thalamic nucleus (POm); all BFP-positive cells were neurons, as evidenced by double labeling with an anti-NeuN antibody (Fig. 4E). Next, we compared the distribution of TVA950-positive neurons with that of the BFP-positive neurons in brain sections, using an anti-HA antibody to immunohistochemically detect the HA-tagged TVA950 protein. Notably, BFP expression was observed only in these thalamic neurons, even though HA-tagged TVA950 protein expression was also observed in cortices with neurons projecting to the injection site of the TVA950-expressing vector, indicating that our vector system selectively visualized a single pathway (Fig. S1A). Most of the TVA950-positive neurons were retrogradely transduced via their axon terminals, because the FuGB2-pseudotyped lentiviral vector showed at least 14.4 times higher preference for transducing neurons via axon terminals over via soma (Fig. S3).

Fig. 4.

Orthogonality of the TVA950–EnvA pair in vivo, demonstrated in rat thalamocortical neurons. (A) Schematic representation of retrograde targeting of TVA950 expression through projection terminals and the selective entry of EnvA vectors from TVA950-expressing neuronal cell bodies. For clarity, EnvA, EnvB, and EnvE are depicted as different shapes. (B) Stereotaxic coordinates of two-step viral injection. First, mutant rabies virus glycoprotein (FuGB2)-pseudotyped TVA950 vectors were injected into primary somatosensory cortex (S1). Three weeks after the first injection, a mixture of Env vectors was injected into the thalamus. (C) Overview of the rat brain sagittal sections counterstained with NeuN antibody. Merged images of NeuN (blue) and BFP (cyan) are shown. BFP-positive cells were restricted to the thalamus (C, Right) and BFP-positive axons innervated S1 (C, Left). (Scale bars: 1,000 μm.) (D) A coronal section near the first injection site. The merged images show the distribution of BFP-positive axons. The boxed area (D, Left) is magnified (D, Right) to show the distribution of BFP-positive axons in the layers of S1. (Scale bars: D, Left, 1,000 μm; D, Right, 250 μm.) (E) A coronal section near the second injection site. The boxed area in E, Left, is magnified in E, Upper Right, to show BFP-positive neurons in the ventral posteromedial (VPM) and posteromedian (POm) thalamic nuclei. The boxed area in E, Upper Right is further magnified in E, Lower Right), to show double labeling of BFP- and NeuN-positive cells. (Scale bars: E, Left, 1,000 μm; E, Upper Right, 250 μm; E, Lower Right, 50 μm.) (F) A representative coronal section stained with anti-HA antibody. The merged image of BFP (cyan) and TVA950–HA (purple) shows that BFP-positive neurons were a subpopulation of TVA950 (HA)-positive neurons. (Scale bar: 250 μm.) (G) Confocal images of a section stained with anti-HA antibody. BFP expression was observed in a subset of TVA950-HA–positive neurons, whereas expression of enhanced GFP (EGFP) or tdTomato expression was not observed. (Scale bar: 50 μm.) Str, striatum; WM, white matter.

A quantitative analysis performed by using sections stained for the HA epitope tag revealed that >99% of BFP-positive neurons were TVA950-positive (ratio of BFP and TVA950 double-positive neurons to all BFP-positive neurons: 1,942/1,959) (Fig. 4F), whereas none were EGFP- or tdTomato-positive neurons (Fig. 4G). Similar results were observed for the other two rats (974/982 and 572/580; Table 1), showing that the EnvA vector leads to specific transduction of TVA950-positive neurons, and other EnvB/EnvE vectors do not transduce TVA950-positive or -negative neurons in vivo.

Table 1.

The orthogonality of the receptor–envelope combinations in vivo

Receptor	Rat no.	No. of sections	No. of BFP+ neurons		No. of EGFP+ neurons		No. of tdTomato+ neurons
			TVA950+	TVA950−	TVA950+	TVA950−	TVA950+	TVA950−
TVA950	A1	5	572	8	0	0	0	0
	A2	5	974	8	0	0	0	0
	A3	5	1,942	17	0	0	0	0
			TVBS3+	TVBS3−	TVBS3+	TVBS3−	TVBS3+	TVBS3−
TVBS3	B1	5	0	0	1,200	12	0	0
	B2	5	0	0	384	13	0	0
	B3	5	0	0	954	10	0	0
			DR-46TVB+	DR-46TVB−	DR-46TVB+	DR-46TVB−	DR-46TVB+	DR-46TVB−
DR-46TVB	D1	5	0	0	0	0	400	2
	D2	5	0	0	0	0	501	1
	D3	5	0	0	0	0	222	0

Env vectors are EnvA–BFP, EnvB–EGFP, and EnvE–tdTomato. Number of BFP-, EGFP-, tdTomato-, and receptor-positive neurons in rats are shown. Note: This table is related to Figs. 4, 5, and 6: Distribution of fluorescent-positive neurons in individual animals.

Orthogonality of the TVB^S3–EnvB Pair in Vivo.

We also tested the specificity of the TVB^S3–EnvB pair in vivo (Fig. 5A). In this experiment, we injected FuGB2–TVB^S3 vector into the S1 region and a mixture of EnvA/EnvB/EnvE vectors into the thalamus 3 wk later (Fig. 5B). The injection coordinates for the FuGB2–TVB^S3 vector were 2.5 mm posterior to those used for the FuGB2–TVA950 vector in the rats shown in Fig. 4. We observed many EGFP-positive axons from thalamic neurons innervating the S1 cortex (Fig. 5C). The EGFP-positive axons projected into cortical layers I, IV, and VI in the S1 cortex near the FuGB2–TVB^S3 vector injection site (Fig. 5D). To compare the distribution of TVB^S3-positive cells with that of EGFP-positive cells, we performed an immunohistochemical analysis using an anti–c-Myc antibody to detect the c-Myc–tagged TVB^S3 protein in brain sections; EGFP-positive cells were detected only in the VPM and POm (Fig. 5E).

Fig. 5.

Orthogonality of the TVB^S3–EnvB pair in vivo, demonstrated in rat thalamocortical neurons. (A) Schematic representation of retrograde targeting of the TVB^S3 expression through projection terminals and selective entry of EnvB vectors from TVB^S3-expressing neuronal cell bodies. (B) Stereotaxic coordinates of two-step viral injection. First, FuGB2-pseudotyped TVB^S3 vectors were injected into S1. Three weeks after the first injection, a mixture of Env vectors was injected into the thalamus. (C) Overview of the rat brain sagittal sections counterstained with NeuN antibody. The merged images of NeuN (blue) and EGFP (green) are shown. EGFP-positive cells were restricted to the thalamus (C, Right), whereas EGFP-positive axons innervated S1 (C, Left). (Scale bars: 1,000 μm.) (D) Coronal section near the first injection site. Merged images show the distribution of EGFP-positive axons. The boxed area (D, Left) is magnified (D, Right) to show the distribution of EGFP-positive axons in the layers of S1. (Scale bars: D, Left, 1,000 μm; D, Right, 250 μm.) (E) Coronal section near the second injection site. The boxed area in E, Left, is magnified in E, Right, to show EGFP-positive neurons in the VPM and POm thalamic nuclei. (Scale bars: E, Left, 1,000 μm; E, Right, 250 μm.) (F) Representative coronal section stained with anti–c-Myc antibody. The merged image of EGFP and TVB^S3–c-Myc (purple) shows that EGFP-positive neurons were a subpopulation of the TVB^S3 (c-Myc)–positive neurons. (Scale bar: 500 μm.) (G) Confocal images of a section stained with anti–c-Myc antibody. EGFP expression was observed in a subset of TVB^S3–c-Myc–positive neurons, whereas BFP or tdTomato expression was not observed. (Scale bar: 50 μm.)

We then performed a quantitative immunohistochemical analysis using the c-Myc epitope tag and found that 96.7% of EGFP-positive neurons were also TVB^S3-positive (ratio of EGFP and TVB^S3 double-positive neurons to all EGFP-positive neurons: 397/410) (Fig. 5F). By contrast, neither BFP- nor tdTomato-positive neurons were observed (Fig. 5G). Similar results were obtained from the other two rats (1,212/1,224 and 964/974; Table 1). These results confirmed that the EnvB vector leads to specific transduction of TVB^S3-positive neurons, and other EnvA/EnvE vectors do not transduce TVB^S3-positive or -negative neurons in vivo.

Orthogonality of the DR-46TVB–EnvE Pair in Vivo.

We next tested the specificity of the DR-46TVB–EnvE pair in vivo (Fig. 6A). In these experiments, FuGB2–DR-46TVB vector was injected into the S1 region, and a mixture of EnvA/EnvB/EnvE vectors was injected into the thalamus 3 wk later (Fig. 6B). The injection coordinates for the FuGB2–DR-46TVB vector in the rats were 2.5 mm posterior to those used for the FuGB2–TVB^S3 vector in the rats shown in Fig. 5. We observed many tdTomato-positive axons from thalamic neurons innervating the S1 cortex (Fig. 6C). Many tdTomato-positive axons projected into S1 cortical layers I, IV, and VI near the FuGB2–DR-46TVB vector injection site (Fig. 6D). To compare the distribution of the DR-46TVB–positive cells with that of tdTomato-positive cells in brain sections, we performed an immunohistochemical analysis using an anti-FLAG antibody to detect the 3× FLAG-tagged DR-46TVB protein. The tdTomato-positive cells were detected only in the VPM and POm (Fig. 6E).

Fig. 6.

Orthogonality of the DR-46TVB–EnvE pair in vivo, demonstrated in rat thalamocortical neurons. (A) Schematic representation of retrograde targeting of DR-46TVB expression through projection terminals and selective entry of EnvE vectors from DR-46TVB–expressing neuronal cell bodies. (B) Stereotaxic coordinates of two-step viral injection. First, FuGB2-pseudotyped DR-46TVB vectors were injected into S1. Three weeks after the first injection, a mixture of Env vectors was injected into the thalamus. (C) Overview of rat brain sagittal sections counterstained with NeuN antibody. Merged images of NeuN (blue) and tdTomato (red) are shown. tdTomato-positive cells were restricted to the thalamus (C, Right), and tdTomato-positive axons innervated S1 (C, Left). (Scale bars: C, Left, 1,000 μm; C, Right, 250 μm.) (D) Coronal section near the first injection site shows the distribution of tdTomato-positive axons. The boxed area (D, Left) is magnified (D, Right) to show tdTomato-positive axons innervate layers I, IV, and VI in the S1. (Scale bars: D, Left, 1,000 μm; D, Right, 250 μm.) (E) Coronal section near the second injection site shows the distribution of tdTomato-positive neurons in detail. The boxed area (E, Left) is magnified (E, Right) to show tdTomato-positive neurons in the VPM and POm thalamic nuclei. (Scale bars: E, Left, 1,000 μm; E, Upper Right, 250 μm; E, Lower Right, 100 μm.) (F) Representative coronal section stained with anti-FLAG antibody. The merged image of tdTomato (red) and DR-46TVB–3× FLAG (purple) shows that tdTomato-positive neurons were a subpopulation of DR-46TVB (3× FLAG)-positive neurons. (Scale bar: 500 μm.) (G) Confocal images of a section stained with anti-FLAG antibody. tdTomato expression was observed in a subset of DR-46TVB–3×-FLAG–positive neurons, whereas BFP (cyan) or EGFP (green) expression was not observed. (Scale bar: 50 μm.) S2, secondary somatosensory cortex; Rt, reticular thalamic nucleus.

We then performed a quantitative analysis using immunohistochemistry with the 3× FLAG epitope tag and found that 99.9% of tdTomato-positive neurons were also DR-46TVB–positive (ratio of tdTomato and DR-46TVB double-positive neurons to all tdTomato-positive neurons: 502/503) (Fig. 6F), whereas none were BFP- or EGFP-positive (Fig. 6G). Among three rats, the average specificity of the DR-46TVB–EnvE pair was 99.7%, the highest among all three receptor–envelope pairs (vs. 99.1% for TVA950–EnvA and 98.7% for TVB^S3–EnvB; Table 1). These results confirmed that the EnvE vector leads to the specific transduction of DR-46TVB–positive neurons, and other EnvA/EnvB vectors do not transduce DR-46TVB–positive or –negative neurons in vivo.

We also examined the orthogonality of the three ASLV receptor–envelope pairs in corticocortical projection neurons in the primary motor cortex (M1) (Fig. S4). In this experiment, FuGB2 vector encoding the ASLV receptors was first injected to M1, and then a second injection of Env vectors was administered to the contralateral M1. The results showed the orthogonality of the three ASLV receptor–envelope pairs in corticocortical neurons of M1 (Table S2), as well as in thalamocortical neurons (Table 1).

Simultaneous Gene Transfer in Cultured Mammalian Cells.

We next examined whether these three receptor–envelope pairs could be used simultaneously to deliver three different genes into three different cell populations in a single culture. For this purpose, we infected a mixture of EnvA/EnvB/EnvE vectors, each expressing different fluorescent proteins, into an intermingled population of HEK 293T cells in which each cell expressed one of the three receptor proteins (TVA950, TVB^S3, or DR-46TVB) (Fig. 7A). The infected cells expressed only one of the three fluorescent proteins in a mutually exclusive manner, creating a three-color cellular mosaic (Fig. 7B). This result confirmed that the infection specificity of the three receptor–envelope pairs is preserved, even in the presence of nonoptimal receptors and Env vectors in vitro.

Fig. 7.

Simultaneous gene transfer into multiple target cells in vitro. (A) Schematic representation of the lentiviral vector constructs used in this test and experimental outline of the infection of HEK 293T cells with receptor-expressing vector and a mixture of Env vectors. (B) Fluorescence images of an intermingled population of HEK 293T cells expressing one of the three receptors transduced with a mixture of Env vectors expressing BFP, EGFP, and tdTomato. (Scale bar: 10 μm.)

ASLV Receptor–Envelope Pairs Enable Triple Pathway-Specific Gene Transfer in Vivo.

To demonstrate the selective and simultaneous labeling of three neuronal populations—each projecting to different target regions—we injected three types of receptor-expressing vectors into the different cortical areas in S1 and a mixture of Env vectors into the thalamus (Fig. 8 A and B). The injection coordinates for each receptor-expressing vector were the same as those shown in Figs. 4–6. In a confocal tiling of a sagittal section, we distinguished three XFPs in different axons (Fig. 8C). BFP-, EGFP-, and tdTomato-positive axons innervated areas near each injection site. In the thalamus, neurons positive for all three types of fluorescence were present. Although the three Env vectors were injected at the same coordinates, the three types of XFP-positive neurons exhibited different distributions in the thalamus (Fig. 8D). Most of the BFP-positive neurons were medial to the EGFP-positive neurons in the thalamus, and the EGFP-positive neurons were more medial than the tdTomato-positive neurons. Furthermore, individual neurons distinctly expressed BFP, EGFP, or tdTomato in a mutually exclusive manner (Fig. 8 E and F).

Fig. 8.

Fluorescent dissociation of three thalamic neuronal populations, each projecting to different cortical regions, visualized simultaneously with orthogonal receptor–envelope pairs. (A) Schematic representation of the lentiviral vector constructs used in this test. (B) Stereotaxic coordinates of two-step viral injection. First, each retrograde TVA950/TVB^S3/DR-46TVB–expressing vector was injected into the different primary somatosensory cortices. Three weeks after the first injection, a mixture of Env vectors was injected into the thalamus. The locations of the serial sections shown in C–E are depicted as nos. 1–5. (C–E) Representative images of sagittal sections stained with antibodies against the three fluorescent proteins (BFP, cyan; EGFP, green; tdTomato, red) and counterstained with NeuN antibody (white). (C) The sagittal section near the injection sites of the FuGB2 vectors shows each fluorescence-positive axon differentially innervating the S1 regions. (D and E) Serial sagittal sections show the distribution of neurons positive for each type of fluorescence in the thalamus. (E) The boxed area near the second injection site is magnified (E, Inset) to show that the three types of projection neurons are intermingled in the thalamus. (Scale bars: C; D, Upper Left; D, Middle; D, Lower; and E, 1,000 μm; D, Upper Right, and E, Inset, 200 μm.) (F) Confocal images of the boxed area in E, Lower Right. Neurons expressed the three fluorescent proteins in a mutually exclusive manner. (Scale bar: 20 μm.)

In conclusion, we identified three ASLV receptor–envelope pairs that are orthogonal in mammalian cells and rat brains. These pairs could be used simultaneously in single cultures or individual rats to fluorescently label three distinct subgroups of neurons.

Discussion

By experimentally identifying and characterizing ASLV receptor–envelope pairs in mammalian cells, we engineered a novel multitargeted gene-transfer system and conducted a proof-of-concept study demonstrating that this system can genetically dissect intermingled neural connections in rat brains. The high orthogonality of three artificial ASLV receptors and their corresponding ASLV envelopes was key to the success of this system and permitted the ASLV-pseudotyped lentiviral vectors to selectively transduce mammalian cells expressing specific receptors in vitro and in vivo.

We combined FuGB2-pseudotyped lentiviral vector (35) with our receptor–envelope system to identify neural populations projecting to a target region in the brain. This retrograde gene-transfer method was selected based on its high efficiency of retrograde transduction via axon terminals (24, 35). Various types of pseudotyped lentiviral vectors, including FuGB2 vector, were developed to achieve retrograde gene transfer by fusing rabies virus glycoprotein to vesicular stomatitis virus glycoprotein (VSV-G) (25, 35–38). The FuGB- and FuGB2-pseudotyped lentiviral vectors show highly efficient retrograde gene transfer (HiRet series) (25, 35). Furthermore, the FuGC- and FuGE-pseudotyped lentiviral vectors show neuron-specific retrograde gene transfer (NeuRet series), but exhibit lower efficiency than that of the HiRet series (37, 38). Therefore, although the NeuRet series was developed after the HiRet series, FuGB2-pseudotyped lentiviral vector is the system that yields the highest efficiency of retrograde gene transfer. Based on these reports, we selected the FuGB2 vector in this study. Although we believe the three ASLV receptor–envelope pairs would also work with the NeuRet series, in vivo tests are necessary to check the NeuRet efficiency for each experimental design. Importantly, to use NeuRet series instead of HiRet, it is necessary to optimize the titers and the waiting time between the injection of NeuRet and Env vectors beforehand.

Our system can also be combined with other retrograde gene-transfer methods, such as recombinant rabies virus (5, 27, 39–43) and adeno-associated virus (AAV) vectors (44–47). However, ASLV pseudotyping may change the pH tolerance and stability of viral particles (48–50); consequently, the infectious titer of viral vectors should be checked before use. Reports about retrograde transduction by AAV vector appear inconsistent; it has been reported as weak or ineffective in some cases (51, 52) and effective in others (44, 45). In addition to this inconsistency, AAV serotypes 8 and 9, which are known to undergo retrograde axonal transport, also anterogradely transduce second-order neurons in vivo (53). This anterograde transport of AAV vector is not suitable for pathway-specific gene expression. Related, the HiRet or NeuRet series selectively undergo retrograde, but not anterograde, axonal transport (24, 25, 36–38).

ASLVs are divided into six viral subgroups, designated A–E and J, based on their cellular receptors (11). From these ASLV subgroups and receptors, we screened for specificity and efficiency using HEK 293T cells and rat brains. In our in vitro assays, the infection rate of the TVB^T–EnvB pair was lower than that of TVB^T–EnvE pair, but much higher than that of control (Fig. 3). However, TVB^T is reported to permit entry of retroviral vector pseudotyped with EnvE, but not EnvB (15). Because we used EnvB-pseudotyped lentiviral vectors at high MOI (MOI = 100), weak receptor–envelope interaction was perhaps detected in our experimental condition. With the TVC–EnvC pair, EnvC vector transduced TVC-expressing neurons only in the absence of other Env vectors (Fig. S3); the reason for these conflicting results remains unclear. Some viruses, such as HIV and hepatitis C virus, use multiple cell-surface components to enter host cells (10). Hence, we speculate that ASLV(C)-pseudotyped lentiviral vectors might require factors other than TVC that are occupied by the other Env vectors (EnvA, EnvB, and EnvE).

In our system, the main factor that determines the transduction efficiency is the expression level of the receptor that interacts with the Env vectors. To increase the efficiency, a 2- to 3-wk period between the injection of FuGB2 and Env vectors is necessary to achieve sufficient expression of the receptors. To increase receptor expression, two refinements are necessary: codon optimization for mammals and shortening of the receptor sequence. Both refinements are expected to increase protein expression level.

To apply our system into optogenetic tools, the insert length will become a problem. The RNA titer of lentiviral vectors decreases with increasing insert length (54). In addition, the ability of ASLV envelope to pseudotype lentiviral vectors is poor relative to that of VSV-G envelope glycoprotein, as demonstrated by its relatively low physical and infectious titers (49). However, we produced Env vectors containing tdTomato (1,475 bp) at high titers [EnvB–tdTomato 8.8 × 10¹⁰ transduction units (TU)/mL; EnvE–tdTomato 6.15 × 10¹⁰ TU/mL), and they worked well in in vivo experiments (Figs. 6 and 8). Possible candidates of frequently used optogenetic tools include hChR2–EYFP (1,662 bp), eNpHR3.0–EYFP (1,683 bp), ChETA–EYFP (1,662 bp), GCaMP6f (1,353 bp), and ArchT–GFP (1,475 bp) (55–59). In GCaMP6f and ArchT–GFP, because the lengths of their DNA sequences are similar to tdTomato, we could produce Env vectors expressing these tools at sufficiently high titers to be used in practice. In hChR2–, eNpHR3.0–, and ChETA–EYFP, because their DNA sequences are not so much longer than that of tdTomato, the titer of Env vectors expressing them could not decrease so much. We will proceed to the next step to apply our receptor–envelope pairs to the frequently used optogenetic tools.

One advantage of this receptor–envelope system is its compatibility with other gene-expression systems. Combining the Cre/loxP and Tet system (6, 60) with the receptor–envelope system allows modifiers of neural activity—such as channelrhodopsin-2 (61, 62), tetanus toxin (63–65), allatostatin (66), and immunotoxin (67–69)—to be selectively introduced and expressed in a more limited population of cell types. In addition, the combination of new optogenetic tools activated by different light wavelengths (70, 71) with the receptor–envelope system may allow the manipulation of each connection separately and simultaneously. Therefore, when combined with other emerging technologies, the system we describe here should make a powerful contribution to the functional analysis of multiple neural populations in vivo.

Materials and Methods

All experiments were approved by the Institutional Review Committee of the University of Tokyo School of Medicine. Lentiviral vectors were produced by cotransfection of HEK293T cells with four plasmids (Table S3) and titrated as described (25, 26). In in vitro experiments, HEK293T cells were serially infected by the FuGB2 vector and the Env vector (3 d after FuGB2 vector infection). Three days later, infected HEK293T cells were collected and fixed with 4% (wt/vol) paraformaldehyde, and analyzed by using fluorescence microscopy or flow cytometer. In in vivo experiments, ten-week-old rats were serially injected with the FuGB2 vector (somatosensory cortex) and the Env vectors (thalamus, 3 wk after FuGB2 vector injections). Injected rats were then perfused with 4% (wt/vol) paraformaldehyde and their brains were processed for histological analyses as described (26). All data are presented as mean ± SE (SEM). We considered P < 0.01 as significant statistical differences. Additional materials and methods, including procedures for in vitro experiments, are provided in SI Materials and Methods.