Virus-Assisted Mapping of Neural Inputs to a Feeding Center in the Hypothalamus

See allHide authors and affiliations

Science  30 Mar 2001:
Vol. 291, Issue 5513, pp. 2608-2613
DOI: 10.1126/science.1056602


We report the development of a pseudorabies virus that can be used for retrograde tracing from selected neurons. This virus encodes a green fluorescent protein marker and replicates only in neurons that express the Cre recombinase and in neurons in synaptic contact with the originally infected cells. The virus was injected into the arcuate nucleus of mice that express Cre only in those neurons that express neuropeptide Y or the leptin receptor. Sectioning of the brains revealed that these neurons receive inputs from neurons in other regions of the hypothalamus, as well as the amygdala, cortex, and other brain regions. These data suggest that higher cortical centers modulate leptin signaling in the hypothalamus. This method of neural tracing may prove useful in studies of other complex neural circuits.

The decision to initiate feeding is dependent on a variety of motivational, metabolic, and hormonal factors including the plasma level of leptin (1, 2). Leptin is an adipocyte-derived hormone whose effects on food intake are mediated through distinct classes of neurons that express the leptin receptor. These neurons are distributed among several hypothalamic nuclei including the arcuate nucleus (ARC) and the ventromedial (VMH), lateral (LH), and dorsomedial (DMH) nuclei (3). Although neurons in these nuclei are known to receive inputs from other brain regions, the precise nature of these inputs has not been established. Identification of such neural inputs requires that one be able to trace neural connections across one or more synapses.

Pseudorabies virus (PRV) has been used previously to trace neural circuits across multiple synapses (4, 5). After injection of PRV into a number of peripheral sites, including the heart, gastrointestinal tract, and liver, viral antigens can be detected in precisely those brain regions known to innervate these organs. PRV has also been used to trace polysynaptic circuits after direct injection into the central nervous system (CNS) (6). However, because the virus infected many different neurons, no information could be deduced regarding inputs to specific classes of neurons, for example, those expressing a particular neurotransmitter, neuropeptide, or receptor. Here, we report the development of a recombinant PRV that is dependent on a Cre-mediated recombination event for replication and for expression of green fluorescent protein (GFP). This virus was used to map CNS inputs to hypothalamic neurons that play a role in regulating food intake.

The Bartha strain of PRV is an attenuated live vaccine strain that is propagated primarily in a retrograde fashion (opposite to the direction of impulse transmission) along chains of synaptically connected neurons. We constructed a PRV-Bartha strain that was conditional both for replication (in nonmitotic cells) and for expression of the Tau-GFP marker (Fig. 1A). We first introduced a deletion into the endogenous thymidine kinase (tk) gene of a recombinant PRV that already carried an insertion of thelacZ gene in the gG locus (TK-BaBlu) (7). The gG gene is dispensable for replication (8). We next created a bicistronic expression cassette in which the cytomegalovirus (CMV) immediate early promoter was cloned upstream of a Tau-GFP fusion gene followed by an internal ribosome entry site (IRES) immediately 5′ to the herpes simplex virus tk gene (9). ALox-Stop-Lox cassette was inserted between the CMV promoter and the start of the Tau-GFP open reading frame. This stop cassette, which is composed of twoLoxP elements flanking an SV40 polyadenylation signal and 5′ splice donor site, prevents expression of downstream sequences unless it is removed by Cre-mediated recombination (10). Thus, expression of both tk and GFP is conditional and requires a Cre-mediated recombination event. This construct was then used to replace the lacZ insert in the TK-BaBlu virus by homologous recombination. The recombinant virus, which we named Ba2001, was initially identified by the loss of β-galactosidase (β-gal) activity and further characterized by Southern blot analysis (11).

Figure 1

Structure and expression properties of PRV strain Ba2001. (A) A recombinant PRV that is dependent on a Cre-mediated recombination event for expression of TK and GFP was created. The parent virus, PRV-Bartha Blue (BaBlu) carries a deletion in the tk gene. The elements of the targeting construct pG2LTHTK are shown below the region of recombination into thegG region of Bartha Blue. Flanking gG sequence is indicated by the red stippled boxes; lacZ open reading framed is indicated by the blue stippled box. Successful recombination of the targeting cassette replaces lacZ, resulting in a white plaque. (B) Conditional expression of Tau-GFP in tissue culture cells. PK15 cells or TE-Cre cells, a Cre-expressing cell line, were infected with Ba2001. GFP expression was observed only after infection of the Cre-expressing cells (right). (C) Cre-dependent GFP expression in Syn1-Cre mice. Ba2001 was injected intracranially into wild-type C57BL/6 (left panels, hypothalamus) or C57BL/6 Syn1-Cre transgenic mice (right panels, cortex) (30). Tau-GFP was detected by immunofluorescence from antibody to GFP (top panels), and the expression of viral late protein gC was detected using antibodies to gC (bottom panels) (31). Viral replication and expression of GFP were observed in the Syn1-Cre but not in the wild-type mice. Scale bars: left panel, 200 μm; right panel, 100 μm.

Because TK is required for viral DNA synthesis in neurons, Ba2001 would not be expected to replicate in neurons that do not express Cre recombinase (12). Infection of a neuron that expresses Cre would be expected to lead to the removal of the stop signals thereby rendering the virus capable of replicating in other neurons in synaptic contact with the original neuron. These infected neurons should also express GFP. The removal of the stop signals is permanent, allowing the subsequent retrograde infection and labeling of any afferent neurons whether or not they express Cre. Conditional expression of GFP in Ba2001 was confirmed by infecting untransfected PK15 and Cre-expressing TE-Cre fibroblasts (13). GFP fluorescence was never observed in Ba2001-infected cells that did not express Cre but was readily detectable in the infected Cre-expressing cells (Fig. 1B).

Viral replication and GFP expression in vivo were also dependent on Cre-mediated recombination (Fig. 1C). Ba2001 was injected intracranially into wild-type C57BL/6 mice and C57BL/6 transgenic Syn1-Cre mice. Mice from the latter strain express Cre in all neurons because the Cre gene is under the control of a 4.4-kb neuron-specific rat synapsin promoter (14). Five days after infection, significant GFP fluorescence was observed in the cortex of Cre-expressing mice but not in the wild-type mice (Fig. 1C). The GFP-expressing neurons also stained for the gC protein, a late viral protein, expressed only after viral DNA replication (15).

To demonstrate that the stop cassette had been excised in vivo, we injected 2.4 × 106 plaque forming units (PFUs) of Ba2001 into the striatum of wild-type and Syn1-Cre mice and killed the animals 7 days later (16). Lysates were prepared from the contralateral striatum of each animal and used to infect PK15 cells in culture. Individual virus plaques were counted and scored for GFP fluorescence. We recovered 75 fluorescent plaques out of a total of 76 plaques per microliter from the lysates of infected Syn1-Cre mice and no fluorescent plaques (0/7) from infected wild-type mice. This indicates that the virus had undergone Cre-mediated recombination in vivo. Southern blots confirmed that the Lox cassette had been removed in the fluorescent viruses (17).

We next used Ba2001 to trace the connections from neurons in which Cre expression was restricted to neurons that express theneuropeptide Y (Npy) gene. NPY is a 36–amino acid peptide that increases food intake and body weight after intracranial or intrahypothalamic injections (18,19). NPY is abundantly expressed in neurons in a large number of brain regions including the cortex, medulla, olfactory bulb, and hypothalamus. Hypothalamic neurons that express NPY also coexpress the leptin receptor and are believed to play a key role in regulating feeding behavior (20). In these experiments, we used a bacterial artificial chromosome (BAC) transgenic mouse line that carries a modified 150-kb Npy BAC in which the Cre recombinase gene was inserted by homologous recombination into the ATG of the first exon of the Npy gene (Fig. 2B) (21). In the resulting BAC, 30 kb of 5′ upstream sequence and 90 kb of sequences 3′ to theNpy gene flank a Cre-poly(A+) cassette. To confirm that the recombination event occurred only in NPY-Cre–expressing cells, we constructed a tk virus conditional only for GFP expression (Fig. 2A). Because this virus is defective for DNA replication, the expression of Tau-GFP should be restricted to those neurons that express NPY (and Cre). After cortical injections of the virus into NPY-Cre mice, large numbers of NPY-expressing cell bodies (scored using a specific antibody against NPY) were observed proximal to the injection site (Fig. 2C, left). GFP fluorescence was evident only in a subset of these neurons (Fig. 2C, middle). Superimposition of the NPY antibody and GFP images revealed yellow and red, but not green, cells (Fig. 2C, right). If GFP expression were independent of Cre, green cells would have been observed. These data confirmed that, in these mice, GFP expression was restricted to Cre- and NPY-expressing neurons.

Figure 2

Specificity of Cre recombination in NPY-expressing cells in vivo. The in vivo requirement of Crefor excision of the Lox-stop cassette was tested in mice that express Cre only in NPY-expressing neurons. Atk virus that was conditional only for GFP was used to infect a transgenic mouse that carried a 150-kb BAC in whichCre had replaced the Npy coding region. (A) Replication-defective virus Ba2000. This virus is identical to Ba2001 except that the expression cassette does not include the tk gene. As a consequence, this virus cannot replicate in nonmitotic cells but still requires Cre for expression of GFP. After infection of Cre-expressing neurons and deletion of the stop cassette, Ba2000 expresses Tau-GFP but remains replication-defective and is unable to spread to higher order neurons. This strategy allows first-order neurons to be visualized exclusively. (B) A BAC transgenic mouse was developed in which the Npy gene was replaced by Cre. A nuclear localization sequence–Cre open reading frame (NLS-Cre)–polyadenylation sequence [poly(A+)] was inserted 5′ to the ATG in the first exon of the Npygene. This BAC includes 30 kb of 5′ upstream and 120 kb of downstream sequence and leads to expression of Cre in NPY neurons. (C) GFP expression is limited to NPY-expressing neurons. Ba2000 was injected into the somatosensory 2 cortex ofNpy-Cre transgenic mice. Sections were stained for NPY or GFP. An overlay of the NPY and GFP immunofluorescence images (right panel) indicates that GFP expression is only observed in neurons that express NPY (yellow cells). Scale bars, 100 μm.

To map neural inputs to NPY-expressing cells that play a role in regulating feeding behavior, we injected Ba2001 into the arcuate nucleus of the NPY-Cre mice stereotaxically. The arcuate nucleus is the hypothalamic site with the highest level of NPY expression (22). Mice were killed at various times after infection, and the pattern of expression of GFP was examined in serial sections of whole brains. At early times, GFP was expressed in several other hypothalamic nuclei, most notably the VMH and LH (Fig. 3A). This result confirms previous evidence that the arcuate nucleus receives inputs from these other hypothalamic nuclei (23). Retrograde infection by the virus from the hypothalamus was also observed in the medial dorsal nucleus of the thalamus (MD) (Fig. 3B) and the dentate gyrus (DG) (Fig. 3C). Five days after infection, the infection had spread to the piriform cortex (Pir) and the ventral basolateral amygdala (BLV) (Fig. 3D). The later appearance of fluorescence at these sites suggests that they are most likely second- or higher-order neurons. Table 1 summarizes the times at which each of the sites was first visualized and the intensity of fluorescence. At earlier stages of infection, GFP expression is restricted to hypothalamic nuclei. Four days after infection, labeling is present in other sites including the amygdala, bed of stria terminalis (BST), and Pir.

Figure 3

Ba2001 infection of NPY-Cre mice after injection into the arcuate nucleus. Npy-Cretransgenic mice were killed 4 days after injection of 1.1 × 106 PFUs of Ba2001 into the arcuate nucleus. (A) Viral spread within the posterior hypothalamus. GFP was expressed in the following hypothalamic structures as indicated: ARC, arcuate nucleus; DMH, dorsomedial nucleus; LH, lateral hypothalamus; VMH, ventromedial hypothalamus; SOX, supraoptic decussation; 3V, third ventricle. Scale bar, 100 μm. (B) Viral spread to the medial dorsal thalamus (MD). Arrows indicate locations of GFP fluorescent neurons. Scale bar, 25 μm. (C) Viral spread to the dentate gyrus (DG). Scale bar, 25 μm. (D) GFP expression in the piriform cortex (PC) and ventral basolateral amygdala (BLV) 5 days postinfection. Scale bar, 25 μm.

Table 1

Summary of Ba2001 infections in NPY-Cre and ObRb-Cre mice. Data from all of the experiments were compiled and the CNS sites that expressed GFP after infections by Ba2001 are shown. Column 1 lists the anatomic sites where GFP expression was observed in the infected NPY-Cre and/or ObRb-Cre mice. Column 2 shows the time postinfection, in days, when GFP labeling was first observed. Columns 3 and 4 show results from Ba2001 infection time course in NPY-Cre transgenic mice. Columns 5 and 6 show results from Ba2001 infection time course in ObRb-IRES-Cre mice. Frequency, the fraction of mice out of the total that expressed GFP at that site. ND, not done. Intensity rating, +++ indicates uniform site labeling intensity among all mice examined; ++, site labeled in all mice with one or more showing decreased intensity; +, site labeled in some but not all mice; −, no labeling observed in any mice.

View this table:

We also attempted to map the inputs to neurons expressing the leptin receptor (ObRb), which is much less abundant than NPY (Fig. 4A). The large size of the ObRgene precluded the use of the BAC transgenic approach, so we introduced an IRES-Cre cassette by homologous recombination immediately 3′ to theObRb stop codon (Fig. 4A). Productive infections were observed after injection of the virus into the hypothalamus but not after injection into the amygdala, Pir, or cerebellum (17). This is consistent with the known distribution of ObRb in the brain (24).

Figure 4

Ba2001 infection of ObRb-IRES-Cre recombinant mice. A total of 1.3 × 106 PFUs of Ba2001 was injected into the arcuate nucleus of embryonic stem (ES) cell–derived mice that express Cre from an IRES element inserted into the 3′ untranslated region of the ObR gene. Animals were killed after infection at the times indicated. Brains were sectioned and GFP expression was visualized by immunofluorescence from antibody against GFP (B to D). Low power (10×) and high power (40×) images are shown on the left and right panels, respectively. (A) ObRb-IRES-Cre targeting construct. A cassette containing an IRES-NLS-Cre and neo gene, flanked byfrt sites, was inserted immediately 3′ to the stop codon in the last exon of ObRb. This cassette was then used for recombination in ES cells, and positive recombinant ES clones were used to generate the ObRb “knock-in” mouse. (B) Expression of GFP in the hypothalamus 3 days postinfection. At 3 days, GFP was expressed in the arcuate, DMH, LH, and VMH. Higher magnification of the DMH (right panel) revealed a possible connection between two DMH neurons (white arrow). Scale bars: left, 200 μm; right, 25 μm. (C) GFP expression in the limbic and cortical brain regions. GFP expression in the BLA, Pir, and LEnt 5 days postinfection (left). Population of GFP-expressing neurons in the BLA (right). Scale bars: left, 200 μm; right, 25 μm. (D) GFP expression in the retrosplenial cortex 7 days postinfection. GFP-expressing neurons in the granular retrosplenial and agranular cortex. Right panel illustrates cells with the morphology of pyramidal cells (red arrows) and interneurons (white arrow). Scale bars: left, 200 μm; right, 50 μm.

We next injected Ba2001 and fluorescent beads (to localize the injection site) into the arcuate nucleus of the ObR-Cre mice and prepared brain sections at various times after infection. A summary of the brain regions visualized after infection is shown in Table 1. As in the NPY experiments, the lag time between viral injection and GFP expression by the various neurons was used to tentatively assign a hierarchy among the connections. At 72 hours postinfection, GFP-expressing neurons were observed in the arcuate nucleus rostral to the injection site (Fig. 4B, left). Although labeled neurons were observed in the contralateral arcuate nucleus, a greater number of labeled neurons were found in the side that received the injection. GFP-expressing neurons were also identified in the LH, VMH, and DMH. Five days after infection, labeled neurons were evident in several extrahypothalamic sites including the basolateral amygdala (BLA), Pir, and lateral entorhinal cortex (LEnt) (Fig. 4C, left). Higher magnification of infected neurons in the BLA revealed what may be a synaptic connection between neurons, a possibility that can be confirmed using electron microscopy (Fig. 4C, right). Labeled neurons were also observed in the agranular and granular retrosplenial cortex (Fig. 4D). In this region, the morphology of the GFP-expressing cells suggested that both pyramidal neurons (red arrows) and interneurons (white arrow) were infected. Labeled neurons in these brain regions are most likely second-order or higher because they appear only at later stages of infection. Fluorescence was restricted to small, discrete numbers of neurons in each of these nuclei and revealed that the infected cells are not contiguous. These data are consistent with previous results indicating that the virus spreads predominantly in a transneuronal fashion rather than laterally (25).

In addition to the sites shown in Fig. 4, labeling was also observed in several other sites, particularly at later times postinfection (Table 1). Five days after infection, labeling was observed in the perirhinal, entorhinal, and somatosensory 2 cortex. At 7 days, GFP-labeled neurons were visible in the somatosensory 1 and retrosplenial cortex, as well as in the CA1 region of the hippocampus and brainstem nuclei. Although neurons that were first observed at 5 days or later are most likely third-order or higher, further analyses will be necessary to determine the precise number of synapses between these sites and the hypothalamus.

Although there was some overlap between the labeled sites observed for the ObRb-Cre and NPY-Cre mice, more brain areas were labeled in the ObRb-Cre mice. This is consistent with the fact that, in the arcuate nucleus, ObRb is coexpressed with NPY, pro-opiomelanocortin, and possibly other neurotransmitters (18). Thus, in the hypothalamus there are several classes of neurons that express ObR, only one of which coexpresses NPY. The GFP-positive sites that are seen only in the ObR-Cre mice may have resulted from projections to infected ObR neurons that do not express NPY.

In summary, we have used a conditional pseudorabies virus to trace the afferent connections of arcuate neurons that express either an abundant transcript, NPY, or a low-abundance transcript, ObRb. In many cases our data are consistent with previous studies in which the connections of unspecified neurons were mapped. In other instances, previously unknown connections have been visualized. For example, projections from the BLA, Pir, and BST to the hypothalamus have been reported using other tracing methods, but the connections from the retrosplenial cortex and LEnt to the hypothalamus have not been detected previously (26–28). Although our viral tracing experiments revealed a variety of afferent inputs in both NPY- and ObRb-Cre mice, it is also possible that additional inputs exist that have escaped detection.

Although the kinetics of neuronal labeling can be used to establish a hierarchy of connections, the designation of neurons as first-, second-, third-, and fourth-order, and so on, is not yet confirmed. Further studies including three-dimensional reconstruction of these images should reveal the precise nature of the connections to the hypothalamus from these brain regions.

These data indicate that hypothalamic neurons that express the leptin receptor receive inputs from a number of CNS centers and are likely to integrate signals from these other sites before in turn transmitting impulses to a set of as yet unidentified, efferent sites. Modifications to this viral tracing system using PRV strains that can be transmitted in the anterograde direction may also make it possible to trace these efferent connections (29). This method and modifications to it may prove useful in studies of other complex neural circuits.

  • * To whom correspondence should be addressed. E-mail: friedj{at}


View Abstract

Navigate This Article