Systematic evaluation of AAV administration strategies for targeting Phox2b + primary visceral sensory neurons

ABSTRACT

Visceral sensory neurons (VSNs) play an essential role in interoception, but it remains challenging to selectively access them using genetic tools due to their anatomical location and shared molecular markers with central autonomic circuits. Although transgenic models visualizing Phox2b-expressing neurons have been developed to genetically label VSNs, concurrent Phox2b expression in the hindbrain limits their selectivity. To overcome these limitations, we engineered several Phox2b promoter-driven adeno-associated virus (AAV) vectors and evaluated the effects of capsid serotype, delivery route, and injection timing. Using systemic delivery of AAV-PHP.S-Phox2b-Cre, we found that postnatal day 8 is an optimal time window for efficient and selective labeling of VSNs with a relatively low dose. The result contrasts with neonatal administration, which led to extensive ectopic labeling at the same dose. Unexpectedly, lingual submucosal delivery of AAV-retro-Phox2b-Cre, initially intended to label tongue-innervating sensory neurons, also labeled most vagal neurons. These labeled neurons were AAV-retro capsid-specific, because local delivery of AAV-PHP.S-Phox2b-Cre labeled cells only near the injection site. Altogether, this study provides practical parameters for selective genetic access to VSNs using AAV.

GRAPHICAL ABSTRACT

Introduction

Visceral sensory neurons (VSNs) are interoceptive neurons that relay internal physiological states from peripheral organs to the central nervous system (Prescott and Liberles 2022). VSNs are critical for maintaining homeostasis in the respiratory and cardiovascular systems, as well as regulating feeding and other autonomic processes. The cell bodies of the many different VSN subpopulations reside in deep cranial sensory ganglia, while their peripheral terminals are distributed across multiple organs. This anatomic organization makes genetic access to specific VSNs challenging.

VSNs express Phox2b, the homeobox transcription factor for epibranchial placode-derived neurons (Pattyn et al. 1997, 1999; D’Autreaux et al. 2011; Dvoryanchikov et al. 2017) (Ki and Jeong 2024). This means BAC transgenic mouse lines like Phox2b-Cre and Phox2b-FLPo can provide genetic access to these populations, and these lines have been used for anatomic and functional studies of visceral sensory circuits (Scott et al. 2011; Hirsch et al. 2013). Peripheral VSNs are not the only cells that exhibit endogenous Phox2b expression; several hindbrain nuclei, including the nucleus of the solitary tract (NTS), area postrema (AP), retrotrapezoid nucleus, and nucleus ambiguus, also express Phox2b (Pattyn et al. 2000) and are labeled by lines, such as Phox2b-Cre and Phox2b-FLPo. Thus, Phox2b-based transgenic mice alone are insufficient for selective labeling of peripheral VSNs to the exclusion of central Phox2b⁺ populations.

Recombinant adeno-associated viruses (AAVs) offer a flexible alternative to transgenic animal models for genetic targeting. Local injection allows the virus to be delivered directly to the target region, thereby excluding unwanted labeling. AAV vectors are widely used in neurogenetics to manipulate cell populations defined by specific molecular markers. AAV-based targeting of VSNs remains challenging, however, because the ganglia containing VSN cell bodies are located deep within the brain, making direct injection invasive and technically demanding (Han and de Araujo 2021). The limited packaging capacity of AAV also restricts the use of large regulatory elements required for highly selective expression. This means selective VSN targeting cannot be accomplished by promoter choice alone. Here, we suggest the inclusion of capsid tropism, delivery route, developmental stage, and viral dose to achieve the necessary level of selectivity.

Several AAV capsids have been engineered to improve tropism for specific tissues and cell populations. For example, AAV-PHP.S is optimized for robust transduction of the peripheral nervous system after systemic delivery, with limited penetration into the adult brain parenchyma (Chan et al. 2017). This serotype efficiently transduces peripheral sensory ganglia, including the nodose ganglion (Asencor et al. 2022). Another engineered capsid, AAV-retro, undergoes efficient retrograde transport from axon terminals to neuronal somata (Tervo et al. 2016). Injection of AAV-retro into peripheral target organs can selectively label peripheral neurons that innervate the injection site (Chen et al. 2020; Zhao et al. 2024). These properties suggest two different ways to target VSNs. AAV-PHP.S can reach peripheral sensory ganglia after systemic delivery, while AAV-retro can access neurons from their peripheral target organs. With both methods, targeting specificity depends on promoter choice, delivery route, developmental timing, and viral dose.

This study aims to develop Phox2b promoter-driven AAV vectors for VSN targeting and systematically evaluate their performance across different capsids, developmental stages, delivery routes, and viral doses. Using AAV-PHP.S, we examined systemic targeting in adult, postnatal day 0 (P0), and postnatal day 8 (P8) mice, paying close attention to trade-offs between labeling efficiency and anatomic specificity. When we asked whether local lingual AAV-retro delivery allowed spatially restricted access to tongue-innervating Phox2b⁺ sensory neurons, we instead found broad labeling across visceral sensory circuits. A direct comparison of lingual AAV-retro and AAV-PHP.S delivery, however, showed that this redistribution depended on capsid identity. Overall, our findings provide practical guidance for viral targeting of Phox2b⁺ VSNs and help define the conditions affecting labeling efficiency and specificity.

Materials and methods

Animals

All animal procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of Korea University College of Medicine (KOREA-2025-0052) and conducted in accordance with institutional guidelines and national regulations. Rosa26-lsl-Salsa6f mice (#031968, Jackson Laboratory) were used in all experiments. Animals were housed under a 12 h light/dark cycle with ad libitum access to food and water. Adult mice weighed approximately 20 g at the time of viral injection, whereas postnatal mice weighed approximately 1.5 g at P0 and 4 g at P8, as indicated. Mice were randomly assigned to experimental groups, including both males and females.

Molecular cloning of pAAV-Phox2b-Cre

The Phox2b promoter fragment was amplified from mouse genomic DNA using the following primers: forward, 5′- ACG CGT AGG GCC CTT TAT CAA ATA CG −3′; reverse, 5′- GGA TCC TGA AAA GGT TCT GGA TGG CTC A −3′. To generate the pAAV-Phox2b-Cre construct, the Ef1a promoter was excised from pAAV-EF1a-Cre (#55636, Addgene) and replaced with the Phox2b promoter fragment. The backbone vector was digested with MluI and BamHI, and the Phox2b promoter sequence was inserted using T4 ligase (M001S, Enzynomics). The final construct was validated through Sanger sequencing.

Adeno-associated virus production

AAV production was performed in AAV293 T cells via calcium phosphate-mediated triple transfection. Cells were cultured in DMEM containing 10% fetal bovine serum and 1% penicillin-streptomycin. One day before transfection, cells were seeded in 150-mm dishes and allowed to reach 70–80% confluency. For each viral preparation, the transfection cocktail was prepared by mixing the transfer plasmid (pAAV-Phox2b-Cre), pHelper plasmid (#240071, Agilent Technologies), and capsid plasmid at a 1:1:1 mass ratio. Either rAAV2-retro helper (#81070, Addgene) or pUCmini-iCAP-PHP.S (#103006, Addgene) were used as capsid plasmids. The culture medium was replaced 16 h after transfection.

Viral particles were harvested 72 h post-transfection using an optimized aqueous two-phase extraction protocol (Guo et al. 2013). Briefly, cells were detached and released into the culture medium. Then, the resulting mixture was treated with chloroform and 5 M NaCl to release viral particles and remove cellular debris. After centrifugation, the virus-laden aqueous phase was collected and precipitated with polyethylene glycol 8000 (6556-1405, Daejung). The resulting viral pellet was resuspended in 50 mM HEPES buffer and incubated with DNase I (18047019, Thermo Fisher Scientific) and RNase A (EN0531, Thermo Fisher Scientific) at 37°C to degrade residual host genomic DNA and RNA. Subsequently, a final chloroform extraction was conducted to remove any remaining lipids. The aqueous phase containing the purified virus was recovered, concentrated, and buffer-exchanged into phosphate-buffered saline (PBS) using Amicon Ultra-0.5 centrifugal filters (UFC510096, Merck).

Adeno-associated virus titration

Viral titers were quantified using SYBR Green-based qPCR targeting the WPRE regulatory element. Before PCR amplification, viral aliquots were treated with DNase I to degrade nonpackaged DNA. qPCR was performed using SYBR Green qPCR Master Mix (GK10002, Glpbio) and the following primers: WPRE-F (5′-GTG GAT ACG CTG CTT TAA TGC C-3′) and WPRE-R (5′-CTC CTC ATA AAG AGA CAG CAA CCA G-3′). Copy numbers were calculated via interpolation from a standard curve generated using a linearized plasmid.

Viral injection procedures

Retro-orbital injection (adult mice)

The retro-orbital sinuses of anesthetized adult mice were injected with AAV-PHP.S-Phox2b-Cre via a 31-gauge insulin needle at a volume of 100 µL per mouse. Mice were monitored during recovery and returned to their cages once ambulatory.

Subcutaneous injection (postnatal mice)

Postnatal mice at P0 (∼1.5 g) or P8 (∼4 g) were given subcutaneous injections of AAV-PHP.S-Phox2b-Cre into the dorsal cervical region using a 30-gauge insulin syringe at a volume of 10 µL per pup. Following the procedure, the pups were returned to their home cage and monitored for normal behavior.

Tongue injection (adult mice)

The tongues of anesthetized supine adult mice (∼20 g) were gently extended with sterile forceps, and the lingual submucosal region was injected with AAV-retro-Phox2b-Cre. Injections were performed using a pulled glass micropipette connected to a Nanoliter 2020 injector (300704, World Precision Instruments). In total, 10 µL was delivered across five injection sites per animal. Following the procedure, the mice were monitored until fully recovered and then returned to their home cages.

Immunohistochemistry

Immunohistochemistry was performed as previously described (Kim et al. 2024; Cho et al. 2025). Following deep anesthetization, mice were given a transcardial perfusion with 0.1 M PBS, followed by 4% paraformaldehyde (PFA) in PBS for fixation. Tissues were carefully dissected and post-fixed in 4% PFA at 4°C overnight. Intestinal tissues were prepared using the Swiss-rolling technique. The tissues were then cryoprotected via immersion in 30% sucrose in PBS at 4°C until fully equilibrated. Cryoprotected samples were then embedded in optimal cutting temperature compound (Tissue-Tek, Sakura Finetek) and sectioned using a cryostat (CM3050S, Leica). The anterior tongue and brain was cut into 45-μm coronal sections, which were allowed to free-float in PBS. The posterior tongue and ganglia was cut into 12-μm sections, which were then mounted directly onto glass slides (HMA-S9911, Matsunami). For immunostaining, sections were blocked for 1 h at room temperature in PBS containing 0.2% Triton X-100 and 5% normal goat or donkey serum. The slides were then incubated overnight at 4°C in primary antibodies diluted in PBS with 0.2% Triton X-100. After three 10-min washes in PBS with 0.2% Triton X-100, the slides were incubated with the appropriate fluorescent secondary antibody for 2 h at room temperature. Following staining, the free-floating sections were mounted on glass slides and counterstained for nuclei with DAPI (1:5,000; D9542, Millipore Sigma). The sections were cover-slipped with antifade fluorescence mounting medium (ab104135, Abcam), and fluorescent images were acquired using a confocal laser-scanning microscope (K1-Fluo, Nanoscope Systems) at an optical section thickness of 1 μm. Images shown in this study are z-stack projections compiled from all optical sections.

The following primary antibodies were used in this study: anti-PGP9.5 (1:400, GTX109637, GeneTex), anti-tdTomato (1:200, AB8181-200, Sicgen), anti-Krt8 (1:1000, ab53280, Abcam), anti-NeuN (1:1000, ab177487, Abcam), and anti-Phox2b (1:200, sc-376997, Santa Cruz). The following secondary antibodies were used in this study: donkey anti-rabbit Alexa488 (1:1000, A21206, Invitrogen), donkey anti-goat Alexa555 (1:1000, A32816, Invitrogen), and donkey anti-mouse Alexa488 (1:1000, A21202, Invitrogen).

Two-photon calcium imaging and taste stimulation

For functional imaging of peripheral sensory endings, a microfluidic tongue stimulation system (μTongue) was used to deliver tastants to the dorsal tongue surface under controlled conditions (Han et al. 2021). An eight-channel fluidic delivery system (Octaflow II, ALA Scientific) was connected to the microfluidic channel inlet via an eight-port manifold (MPP-8, Harvard Apparatus). Outflow from the device was controlled using a syringe pump to maintain a constant flow rate of approximately 200 μL/min. Proper tastant delivery to fungiform taste buds was periodically verified via perfusion of a fluorescent tracer solution. Acesulfame potassium (AceK, 40 mM) dissolved in distilled water (DW) was used as a sweet taste stimulus, with DW as the vehicle control. To establish a stable baseline, DW was perfused continuously for ≥ 3 min before stimulation. Each taste stimulus was delivered for 20 s, followed by a thorough washout with DW. Two-photon calcium imaging was performed with an IVIM 2P microscope equipped with a 920 nm excitation laser (IVIM-MS2, IVIM Technologies). GCaMP6f fluorescence signals were acquired at 15 Hz and analyzed post hoc using ImageJ.

Post-imaging data processing

Regions of interest (ROIs) corresponding to Phox2b + sensory terminals were manually identified from maximum GCaMP6s fluorescence projections based on their distinctive morphology beneath fungiform papillae (FuP). For each frame, fluorescence intensity (F) was calculated as the mean pixel value within each defined ROI. To correct for background noise, the mean fluorescence of a nearby unlabeled region was subtracted from each ROI signal. Baseline fluorescence (F₀) was defined as the mean intensity during the 15 s pre-stimulus period preceding each taste delivery. The relative fluorescence change was calculated as ΔF/F = (F(t) - F₀) / F₀. AceK was delivered three times, and only units that showed a fluorescence response in at least two of the three trials were included in the analysis.

Results

Development of a recombinant AAV targeting Phox2b visceral sensory neurons

We first examined whether the mouse Phox2b promoter could drive selective reporter expression in Phox2b⁺ visceral chemosensory neurons in vivo. We cloned a PCR-amplified 1,925-bp genomic region upstream of the Phox2b transcription start site into an AAV backbone upstream of the Cre recombinase coding sequence (Figure 1A). We avoided direct ganglion injections by initially constructing this package in the PHP.S capsid, which is optimized for peripheral neuron transduction following systemic administration (Chan et al. 2017).

Figure 1.

Generation and systemic administration of AAV-PHP.S-Phox2b-Cre in adult mice. (A) Schematic showing the AAV-Phox2b-Cre construct. Red, WPRE; green, poly(A). (B) Schematic for RO injection of AAV-PHP.S-Phox2b-Cre (2.2 × 10¹² vg/mouse) into adult Rosa26-lsl-Salsa6f mice (>20 g). (C–G) Representative confocal images showing tdTomato fluorescence in the indicated organs. (C) Tongue, including fungiform papillae (FuP) and circumvallate papillae (CVP) with anti-PGP9.5 (green, neurons) and tdTomato (red). (D) Tongue. Anti-Krt8 (green, taste buds) and tdTomato (red). (E) Small intestine. Anti-PGP9.5 (green) and tdTomato (red). (F) Nodose-petrosal ganglion (NPG). Anti-NeuN (green, neurons) and tdTomato (red). (G) Hindbrain showing the rostral (bregma, −6.72 mm) and caudal (bregma, −7.55 mm) nucleus solitary tract (NTS). Anti-NeuN (green) and tdTomato (red); dashed lines delineate the NTS, Sp5 (spinal trigeminal tract), Sp5I (spinal trigeminal nucleus, interpolar part), Sp5C (spinal trigeminal nucleus, caudal part), AP (area postrema), DMX (dorsal motor nucleus of the vagus), and XII (hypoglossal nucleus). Scale bars: 50 μm (C, D), 100 μm (E, F), and 500 μm (G). n = 3 mice.

To assess the resulting targeting profile, we injected AAV-PHP.S-Phox2b-Cre into the RO sinus of Rosa26-lsl-Salsa6f mice at a high titer (>10¹² vg/mouse) and examined transduction patterns 2 weeks later (Figure 1B). We detected reporter expression in visceral tissues extending from the oral cavity to the gastrointestinal tract. Within the gustatory system, we found strongly tdTomato-positive axonal fibers innervating the Krt8⁺ taste buds of fungiform (FuP) and circumvallate papillae (CVP) (Figure 1C & D). In the small intestine, we observed tdTomato expression in axonal fibers co-labeled with the pan-neuronal marker PGP9.5 (Figure 1E). In the nodose-petrosal ganglion complex (NPG), which contains glossopharyngeal and vagal sensory somata, we detected tdTomato expression in a subset of NeuN-positive neurons (Figure 1F).

Next, we examined the central projections of these neurons in the brainstem. We observed dense tdTomato labeling of axonal arbors throughout the NTS, including both the rostral gustatory and caudal general visceral divisions (Figure 1G). In contrast, the spinal trigeminal tract and interpolar spinal trigeminal nucleus (Sp5I), which both receive somatosensory inputs, showed only minimal labeling (Figure 1G). Thus, unlike the broader distribution previously reported with the AAV-PHP.S-CAG vector (Asencor et al. 2022), we found AAV-PHP.S-Phox2b-Cre labeling was largely restricted to visceral sensory pathways. We also observed tdTomato-positive cell bodies in the AP (Figure 1G), which is consistent with endogenous Phox2b expression in hindbrain nuclei (Pattyn et al. 2000). The AP is a circumventricular organ with a relatively permeable blood-brain barrier, which is likely what allowed some local somatic labeling despite the limited brain penetration of PHP.S in adult mice. These results demonstrate that the 1,925-bp Phox2b promoter is sufficient to confer selective transgene expression in Phox2b-expressing VSNs following AAV delivery.

Systemic AAV-PHP.S transduction efficiency in adults was dose-dependent

After confirming that systemic delivery of AAV-PHP.S-Phox2b-Cre effectively labeled the targeted VSNs of adult mice at a relatively high viral load (>10¹² vg/mouse), we next asked whether a lower dose could produce a comparable result. To test this, we similarly injected adult Rosa26-lsl-Salsa6f mice with a ten-fold lower dose of AAV-PHP.S-Phox2b-Cre (2.2 X 10¹¹ vg/mouse) and evaluated reporter expression 2 weeks later (Figure 2A).

Figure 2.

Subthreshold systemic administration of AAV-PHP.S-Phox2b-Cre in adult mice. (A) Schematic for RO injection of AAV-PHP.S-Phox2b-Cre (2.2 × 10¹¹ vg/mouse) into adult Rosa26-lsl-Salsa6f mice (>20 g). (B–F) Representative confocal images showing tdTomato fluorescence in the indicated organs. (B) Tongue. Anti-PGP9.5 (green) and tdTomato (red). (C) Tongue. Anti-Krt8 (green) and tdTomato (red). (D) Small intestine. Anti-PGP9.5 (green) and tdTomato (red). (E) NPG. Anti-NeuN (green, neurons) and tdTomato (red). (F) Hindbrain, showing the rostral (bregma, −6.72 mm) and caudal (bregma, −7.55 mm) NTS. Anti-NeuN (green) and tdTomato (red). Scale bars: 50 μm (B, C), 100 μm (D, E), and 500 μm (F). n = 3 mice.

Unfortunately, at this lower viral load, we could not detect any tdTomato expression in peripheral target organs, including CVP and FuP taste buds, intestinal sensory neurons, or in sensory somata in the NPG (Figure 2B–E). Nor could we detect labeled axonal arborization in either the rostral or caudal divisions of the NTS (Figure 2F). The only signal we could detect was in the AP, but even this was sparse without a clear somatic pattern (Figure 2F). Thus, efficient systemic labeling in adult mice requires a viral dose above 10¹² vg.

P0 systemic AAV-PHP.S delivery increased transduction efficiency but compromised specificity

Although 2.2 X 10¹¹ vg was insufficient in adults, we asked whether a similar low dose would produce robust transduction in neonatal mice, which have a much lower body mass (∼1.5 g). To test this, we injected P0 Rosa26-lsl-Salsa6f mice subcutaneously with AAV-PHP.S-Phox2b-Cre (1.68 X 10¹¹ vg) and analyzed reporter expression 8 weeks later when the mice had reached adulthood (Figure 3A).

Figure 3.

Systemic administration of AAV-PHP.S-Phox2b-Cre at P0. (A) Schematic for subcutaneous injection of AAV-PHP.S-Phox2b-Cre (1.68 × 10¹¹ vg/mouse) into P0 neonates (<1.5 g). (B–H) Representative confocal images showing tdTomato fluorescence in the indicated organs. (B) Tongue. Anti-PGP9.5 (green) and tdTomato (red). (C) Tongues. Anti-Krt8 (green) and tdTomato (red). (D) Small intestine. Anti-PGP9.5 (green) and tdTomato (red). (E) NPG. Anti-NeuN (green) and tdTomato (red). (F) Hindbrain showing the rostral (bregma, −7.15 mm) and caudal (bregma, −7.55 mm) NTS. Anti- NeuN (green) and tdTomato (red). (G) Choroid plexus sections. Anti-Phox2b (green) and tdTomato (red). (H) Forebrain (bregma, 1.42 mm); dashed lines delineate the cortex and subcortex, with a magnified view of ectopic tdTomato expression in the cortical region. Scale bars: 50 μm (B, C), 100 μm (D, E), and 500 μm (F–H). n = 3 mice.

We found P0 administration led to widespread tdTomato labeling across the peripheral and central components of the visceral sensory pathway, including dense tdTomato-positive innervation in both FuP and CVP taste buds (Figure 3B & C) and in sensory neurons of the small intestine (Figure 3D). Most NeuN-positive neurons in the NPG were labeled (Figure 3E), indicating greater transduction efficiency than that observed after high-dose systemic AAV delivery in adults.

This increased efficiency, however, was accompanied by a loss of targeting specificity in the brain. We observed dense axonal labeling in the rostral and caudal NTS and in the AP, but also in the Sp5 (Figure 3F), which remained unlabeled in our experiments with adult systemic delivery. We also found robust reporter expression in the choroid plexus, where tdTomato labeling overlapped with anti-Phox2b immunoreactivity (Figure 3G), confirming co-expression with endogenous Phox2b. Notably, there were also tdTomato-positive cells in several forebrain regions, including the cerebral cortex (Figure 3H). Thus, although P0 administration markedly increased transduction efficiency, it came at the expense of targeting specificity.

Systemic AAV-PHP.S delivery at postnatal day 8 improved specificity while maintaining efficiency

Given that the mouse BBB reaches structural maturity around P7 and that P8 mice still have a relatively low body mass (∼4 g), we next asked whether P8 delivery would retain efficient transduction at a relatively low viral load while also reducing nonspecific ectopic labeling. We administered subcutaneous injections of AAV-PHP.S-Phox2b-Cre (4.28 × 10¹¹ vg) to P8 Rosa26-lsl-Salsa6f mice and then evaluated reporter expression once the mice reached adulthood (Figure 4A).

Figure 4.

Systemic administration of AAV-PHP.S-Phox2b-Cre at P8. (A) Schematic for subcutaneous injection of AAV-PHP.S-Phox2b-Cre (4.28 × 10¹¹ vg/mice) into P8 mice weighing 1.5∼4.0 g. (B–H) Representative confocal images showing tdTomato fluorescence in the indicated organs. (B) Tongue. Anti-PGP9.5 (green) and tdTomato (red), including a magnified view of the taste buds and dense sensory nerves. (C) Tongue. Anti-Krt8 (green) and tdTomato (red). (D) Small intestine. Anti-PGP9.5 (green) and tdTomato (red). (E) Whole-mount NPG. Anti-NeuN (green) and tdTomato (red). (F) Trigeminal ganglion (TG). Anti-NeuN (green) and tdTomato (red). (G) Hindbrain, showing the rostral (bregma, −6.72 mm) and caudal (bregma, −7.55 mm) NTS. Anti-NeuN (green) and tdTomato (red). (H) Forebrain (bregma, 1.98 mm). Anti-NeuN (green) and tdTomato (red). Scale bars: 50 μm (B, C), 100 μm (D–F), and 500 μm (G, H). n = 3 mice.

In contrast to our results at P0, we found mice injected at P8 retained strong labeling of visceral target organs while also showing improved specificity. In peripheral organs, the dense innervation we observed in mice injected at P0 was largely preserved, with tdTomato-positive fibers innervating FuP and CVP taste buds (Figure 4B, C) and intestinal villi (Figure 4D). In the NPG, most neurons expressed tdTomato (Figure 4E), indicating efficient genetic access to vagal sensory somata. At the same time, however, there were far fewer tdTomato-expressing neurons in the trigeminal ganglion (TG) than in the NPG (Figure 4F).

In the brainstem, although central labeling remained prominent in the rostral and caudal NTS and the AP, there was less labeling in the Sp5 than in mice injected at P0 (Figure 4G). In addition, the widespread ectopic reporter expression in forebrain regions associated with P0 injection was largely absent in mice injected on P8 (Figure 4H). Thus, systemic P8 delivery of a relatively low viral dose (∼10¹¹ vg) preserves efficient labeling of Phox2b⁺ VSNs but improves specificity compared to P0 delivery.

Lingual submucosal AAV-retro delivery enhances broad labeling of VSNs

Thus far, we found systemic PHP.S administration efficiently labels Phox2b⁺ VSNs but lacks the anatomic resolution required to isolate neurons innervating a specific target organ. To achieve organ-specific labeling, we next sought to leverage the retrograde transport properties of the AAV-retro serotype (Tervo et al. 2016). We hypothesized that the combination of the Phox2b promoter with local AAV-retro delivery should enable selective labeling of sensory neurons projecting to the injected tissue. To evaluate this approach, we generated AAV-retro-Phox2b-Cre and injected it into the lingual submucosa of adult Rosa26-lsl-Salsa6f mice to target tongue-innervating Phox2b⁺ sensory neurons (Figure 5A), particularly the gustatory afferents innervating taste buds.

Figure 5.

Submucosal lingual administration of AAV-retro-Phox2b-Cre in adult mice. (A) Schematic for lingual submucosal administration of AAV-retro-Phox2b-Cre (2.2 × 10¹¹ vg/mice) in adult mice (>20 g). (B–H) Representative confocal images showing tdTomato fluorescence in the indicated organs. (B) Tongue. Anti-PGP9.5 (green) and tdTomato (red). (C) Tongue. Anti-Krt8 (green) and tdTomato (red). (D) Whole-mount NPG. Anti-NeuN (green) and tdTomato (red). (E) Small intestine. Anti-PGP9.5 (green) and tdTomato (red). (F) TG. Anti-NeuN (green) and tdTomato (red). (G) Hindbrain, showing the rostral (bregma, −6.95 mm) and caudal (bregma, −7.55 mm) NTS. Anti-NeuN (green) and tdTomato (red). (H) Forebrain (bregma, 1.98 mm). Anti-NeuN (green) and tdTomato (red). Scale bars: 50 μm (B, C), 100 μm (D–F), and 500 μm (G, H). n = 3 mice.

While we did find tdTomato labeling in gustatory neurons innervating FuP and CVP (Figure 5B, C), we unexpectedly also found tdTomato labeling in most NPG neurons (Figure 5D), as well as in vagal sensory axons in the intestine (Figure 5E). This pattern was like what we observed with high-titer systemic AAV-PHP.S delivery. In the brainstem, labeled central terminals spanned the rostral gustatory NTS and extended into the caudal NTS (Figure 5G), indicating transduction beyond the initial injection site.

Despite this broad labeling, targeting specificity was largely preserved. We did not detect any tdTomato signal in the TG (Figure 5E), indicating an exclusion of somatic sensory neurons. Nor did we detect any labeling of the spinal trigeminal nucleus (Sp5) or the forebrain (Figure 5G). Unlike with P0 systemic injection, lingual AAV-retro delivery broadly labeled VSNs without any detectable ectopic labeling.

AAV-retro’s broad labeling of the lingual submucosa is capsid-specific

The unexpected systemic labeling associated with lingual AAV-retro injection suggested that this phenomenon was either a general consequence of mucosal delivery or a specific property of the AAV-retro capsid. To determine which, we injected the tongue of adult Rosa26-lsl-Salsa6f mice with a comparable dose (3.77 × 10¹¹ vg/mouse) of AAV-PHP.S-Phox2b-Cre (Figure 6A).

Figure 6.

Submucosal lingual administration of AAV-PHP.S-Phox2b-Cre in adult mice. (A) Schematic for lingual submucosal administration of AAV-PHP.S-Phox2b-Cre (3.77 × 10¹¹ vg/mouse) via five injection sites in adult mice (>20 g). (B–H) Representative confocal images showing tdTomato fluorescence in the indicated tissues. (B) Tongue. Anti-PGP9.5 (green) and tdTomato (red). (C) Tongue. Anti-Krt8 (green) and tdTomato (red). (D) Tongue. Anti-S100 (green) and tdTomato (red). (E) Small intestine. Anti-PGP9.5 (green) and tdTomato (red). (F) Whole-mount NPG. Anti-NeuN (green) and tdTomato (red). (G) Sectioned TG. Anti-NeuN (green) and tdTomato (red). (H) Hindbrain showing the rostral (bregma, −6.72 mm) and caudal (bregma, −7.55 mm) NTS. Anti-NeuN (green) and tdTomato (red). (I) Forebrain (bregma, 1.98 mm). Anti-NeuN (green) and tdTomato (red). Scale bars: 50 μm (B, C), 100 μm (D–F), and 500 μm (G, H). n = 3 mice.

In contrast to the AAV-retro delivery, this localized PHP.S delivery produced a highly restricted transduction pattern. In the tongue, we observed robust tdTomato expression in sensory neurons innervating FuP, with terminal boutons clearly visible within the taste buds (Figure 6B, C). In contrast, we observed only sparse sensory innervation of the CVP (Figure 6B, C). PHP.S delivery also produced dense reporter expression in the underlying lingual musculature and connective tissue. We were able to attribute some of this signal to Schwann cells by confirming co-labeling with anti-S100 (Figure 6D), but there was clear labeling of several other cell types as well. This finding suggests high local PHP.S concentrations can cause nonspecific transduction of nonneuronal cells under these conditions.

In contrast to the AAV-retro group, we did not observe any tdTomato labeling from distant visceral organs, including the small intestine, nor did we detect any tdTomato-positive somata in the NPG or TG (Figure 6E, F). In the brainstem, we found no labeling in the caudal NTS but robust labeling in the rostral NTS. Since the rostral NTS is the canonical target of FuP-specific gustatory afferents (Figure 6G), this result is consistent with the tongue labeling pattern. We also observed robust labeling in the hypoglossal nucleus (12N), which contains Phox2b⁺ motor neurons that innervate the lingual musculature (Pattyn et al. 2000). Neither the Sp5 nor the forebrain exhibited any labeling (Figure 6G, H).

Since local AAV-PHP.S delivery did not show the broad peripheral redistribution associated with AAV-retro, we conclude that redistribution across the visceral sensory axis depends on the viral capsid.

Functional responses of Phox2b⁺ gustatory neurons

To assess the suitability of the viral strategies investigated in this study for functional interrogation of Phox2b⁺ sensory neurons, we performed in vivo two-photon calcium imaging of the tongue under three conditions (Figure 7): adult high-dose delivery of AAV-PHP.S-Phox2b-Cre (systemic), P0 low-dose delivery of AAV-PHP.S-Phox2b-Cre (systemic), and adult delivery of AAV-retro-Phox2b-Cre (lingual). In each group, we used Rosa26-LSL-Salsa6f mice to express the ratiometric calcium indicator GCaMP6f in Phox2b⁺ neurons. Using a microfluidic tongue perfusion system to control stimulus delivery (Han et al. 2021), we recorded GCaMP6f signals from Phox2b⁺ sensory terminals in the lingual epithelium. Across all three conditions, oral AceK application reliably evoked calcium transients tightly coupled to stimulus onset (Figure 7). We found qualitatively similar response profiles across all three delivery methods, indicating that all three strategies provide functional access to taste bud–innervating Phox2b + sensory terminals. Therefore, all three viral strategies supported in vivo physiological imaging of gustatory afferents.

Figure 7.

Functional validation of Phox2b + gustatory neurons across viral delivery strategies. (A) Representative GCaMP6f images showing the baseline (pre-stimulus), peak (during stimulus), and decay (post-stimulus) phases for each of the three viral delivery strategies: adult systemic high-dose AAV-PHP.S-Phox2b-Cre (top), P0 systemic low-dose AAV-PHP.S-Phox2b-Cre (middle), and adult lingual AAV-retro-Phox2b-Cre (bottom). n = 3 mice per group. (B) Representative calcium transient traces (ΔF/F) in response to 40 mM AceK. Colored boxes indicate the 20-s stimulus delivery period.

Discussion

In this study, we produced Phox2b promoter-driven AAV vectors and systematically compared multiple viral delivery strategies for the targeting of VSNs. After comparing capsid serotype, delivery route, developmental timing, and viral dose, we found that promoter selectivity alone was insufficient to predict targeting outcomes. Rather, we found that various combinations of these parameters produced significant differences in labeling efficiency and anatomic specificity.

Systemic AAV-PHP.S delivery in adult mice requires a relatively high viral dose (Chan et al. 2017; Asencor et al. 2022). We therefore asked whether earlier administration could achieve efficient labeling with less virus. At P0, we were able to achieve robust VSN labeling with a dose that was insufficient in adults. This increased efficiency, however, was accompanied by ectopic labeling in the cerebral cortex and other somatosensory regions. We suspect this broad labeling reflects the immaturity of the BBB at birth, although developmental changes in Phox2b promoter activity may also contribute. With delivery at P8, we observed strong VSN labeling accompanied by a significant reduction in ectopic labeling.

Lingual submucosal delivery of AAV-retro-Phox2b-Cre produced an unexpected labeling pattern. Although this injection was intended to target Phox2b⁺ sensory afferents innervating the tongue, we also found labeling in most NPG neurons and intestinal vagal sensory projections. Not only is this difficult to explain by simple local spread from the lingual injection site, the broader labeling seemed somewhat specific, with TG neurons and somatosensory brainstem regions showing little or no reporter expression. Thus, although AAV-retro produced labeling far beyond the local injection site, the Phox2b promoter continued to restrict expression largely to visceral sensory populations. We suspect viral entry to the lingual target field enables access to convergent or closely apposed visceral sensory pathways. The mechanism remains unclear, but the absence of similar labeling following lingual AAV-PHP.S delivery suggests this pattern is specific to AAV-retro rather than a consequence of submucosal injection.

Lingual AAV-retro delivery produced broad VSN labeling even at the lowest viral dose tested. Although labeling was not restricted to the injected organ, our approach allowed us to avoid direct ganglion injection while also maintaining selectivity for Phox2b⁺ VSNs. It may be possible to improve organ-specific targeting by further reducing the viral dose, but this remains to be tested.

Although this study used Phox2b as a molecular marker for VSNs, a similar strategy could be applied to other compact promoters or enhancers that mark peripheral neuronal populations with somata in anatomically inaccessible ganglia or plexuses (e.g. sympathetic and parasympathetic postganglionic and enteric neurons). When combined with AAV vectors carrying Cre-dependent expression constructs, the viral strategies we employed may also enable minimally invasive access to deep peripheral neuronal populations in nontransgenic animal models.

In conclusion, selective viral access to VSNs depends on promoter specificity, capsid serotype, delivery route, developmental timing, and viral dose. All these parameters should be considered when designing viral strategies for genetic access to Phox2b⁺ VSNs and other deep peripheral neuronal populations.

Funding Statement

This work was supported by a National Research Foundation of Korea (NRF) grant, funded by the Korean Government (Ministry of Science and ICT) [grant number RS-2023-00208193 and RS-2024-00406845 to Y.T.J.], and the Korea Health Technology R&D Project, administered by the Korea Health Industry Development Institute (KHIDI) funded by the Ministry of Health & Welfare [grant number RS-2024-00403511 to D-H.K.].

Disclosure statement

No potential conflict of interest was reported by the author(s).

Declaration of generative AI use

The authors report generative AI was not used in their research or preparation of this manuscript.

Material availability

All the materials used in this study are available from the corresponding author upon reasonable request.