Intravenous rAAV9 Produces Time-Resolved Parenchymal Labeling Downstream of the Vasculature in Adult Mice

TL;DR

This study maps which cells outside blood vessels in adult mice become labeled after an intravenous gene therapy injection, helping improve understanding of how genetic material reaches tissues.

Contribution

The study provides a time-resolved map of parenchymal labeling after intravenous rAAV9 delivery in adult mice.

Findings

Durable labeling was observed in several parenchymal cell populations across multiple organs and the nervous system.

Reporter activation was consistent in nonvascular cells relevant to systemic gene delivery.

The findings suggest potential mechanistic routes like viral transcytosis or extracellular vesicle transfer.

Abstract

Systemic gene therapy is often limited by uncertainty about which cells outside blood vessels actually receive and express delivered genetic cargo after injection into the bloodstream. In this study, we injected a single dose of a gene-delivery virus into healthy adult mice and used a fluorescent reporter system that permanently marks cells that receive a genetic “switch” carried by the virus. By examining many organs over time, we created a time-resolved map of which cell types outside the vasculature become labeled after an intravenous dose. We found durable labeling in several parenchymal cell populations across peripheral organs and the nervous system, identifying specific tissues that warrant mechanistic follow-up to determine how labeling occurs after vascular exposure. These data provide a reference resource for researchers optimizing minimally invasive delivery strategies and for interpreting downstream labeling events after systemic administration.

Intravenous delivery of recombinant adeno-associated virus serotype 9 can lead to reporter activation in cell types beyond the vasculature, but the routes enabling downstream parenchymal labeling remain unclear. Here, we provide a systematic, time-resolved map of parenchymal labeling after a single intravenous dose of rAAV9 encoding Cre recombinase under a ubiquitous promoter in healthy adult Ai9 reporter mice. Following retro-orbital administration, we quantified tdTomato-positive labeling across 25 targets at multiple time points over six months and observed durable reporter activation in several nonvascular parenchymal populations relevant to systemic gene-delivery applications. We also identify a set of parenchymal cell types that are consistently labeled in both this vascularly initiated reporter system and our prior adult VE-cadherin-driven reporter paradigm, supporting a connection to vascular exposure without asserting lineage relationships. These results nominate mechanistic routes for future disambiguation, including viral transcytosis across endothelium, endothelial cell transdifferentiation and extracellular-vesicle-mediated transfer. The dataset and methods provide a reference framework for investigators optimizing systemic delivery and interpreting downstream labeling in vivo.

Figures

Full text

1. Introduction

Genetic therapy for human disease requires the directed delivery of gene editing tools to the cells of interest. This limitation is highly problematic for most cells and often results in a brute force non-specific infusion into the parenchyma with low target efficiency and high off-target cell editing [1]. Interestingly, there is mounting evidence that lowly invasive genetic therapy is available for some difficult-to-target cells by editing cells within the vascular compartment via intravenous injection [2]. The simple, direct targeting of gene therapy agents, such as recombinant adeno-associated virus (rAAV), to endothelial cells and circulating cells has shown some efficacy in improving outcomes in numerous preclinical animal models [3,4,5,6]. However, the genetic editing of parenchymal cells is conceptually challenging, given that viral, protein, and nucleic acid-based gene therapy agents introduced into the vasculature must pass through endothelial cells and an extensive extracellular matrix to reach their target. Yet, empirical evidence has demonstrated that the intravenous delivery of rAAV in mammalian organisms results in highly efficient genetic editing of cardiac myocytes [7,8] and other parenchymal cell types, with some at greater than 90% efficiency [9]. Importantly, the question of how this occurs has never been convincingly addressed. Therefore, fully harnessing this process for maximal therapeutic benefit requires the identification of the diverse cell types that are edited downstream of the vasculature. Likewise, proper evaluation of the potential pitfalls requires us to understand the mechanism of the intravenous genetic editing of parenchymal cell types.

Several observations raise the possibility that endothelial cells could contribute—directly or indirectly—to some downstream parenchymal labeling after intravenous rAAV exposure; however, endothelial lineage relationships are not established by current tracing approaches. Three possible mechanisms for transferring activated gene-editing machinery to downstream cell types are cell transdifferentiation, extracellular vesicle (EV) emission and transcytosis of viral particles. Whether endothelial cells undergo transdifferentiation in adults is a central question that could have significant implications for gene therapy and disease processes. Growing evidence supports the notion of endothelial cell plasticity. It has been reported that primary mouse neural precursor cells spontaneously differentiate into endothelial cells at an appreciable rate [10]. The predisposition for endothelial cells to undergo transdifferentiation is also believed to be an important feature for several forms of vascular disease [11,12,13]. Tang et al. reported that, following vascular injury, specialized endothelial cells differentiate into chondrogenic and smooth muscle cells [14]. This group also isolated a population of multipotent endothelial cells from adult mice that readily differentiated into several distinct cell types in vitro [14]. Likewise, endothelial cell EV transmission to cell types outside of the vasculature is a well-studied phenomenon that has been linked to numerous homeostatic processes and disease pathogenesis [15,16,17]. Transcytosis of viral capsids has also been reported [18,19,20] and could contribute to labeling, although this process does not explain parenchymal cell tracing from vascular endothelial cadherin (VEcad) promoter-driven labeling [21].

Our previous work [21] reported VEcad-promoter-driven labeling patterns that were consistent with, but did not prove, an endothelial contribution to certain downstream labeling events using an endothelial cell promoter-driven (VEcad) model of cell lineage tracing in adult mice. However, this VEcad promoter transgenic mouse line was shown to have rare non-endothelial cell expression, thus confounding interpretation. We aim to build upon this work by performing a directly comparable study utilizing intravenous delivery of rAAV to initiate cell lineage tracing in the same transgenic mouse reporter line, Cg-Gt(ROSA)26Sortm9(CAG-tdTomato)Hze) (Ai9). Cell types labeled in both the VEcad-promoter model and the intravenous rAAV-Cre paradigm are priorities for mechanistic follow-up to test whether a vascular exposure step is involved and, if so, whether mechanisms such as EV-associated transfer or rare cell-fate changes contribute. In addition, we wish to provide a comprehensive study regarding the identification of parenchymal cells edited by intravenous rAAV9 by examining 25 different tissues at multiple time points over a six-month period.

Conceptually, VEcad-promoter lineage tracing and intravenous rAAV-Cre tracing interrogate different constraints on the same putative vascular-to-parenchymal labeling axis. VEcad-promoter (CreER) models label cells defined by endogenous VEcad promoter activity (primarily endothelium) after tamoxifen induction, enabling lineage relationships but remaining sensitive to promoter leak/off-target expression and induction parameters. By contrast, intravenous rAAV9-CMV-Cre labels cells based on viral biodistribution, vascular exposure, transduction efficiency, and promoter activity, providing broad access to the vasculature but permitting alternative routes such as transcytosis or EV-associated secondary spread. Accordingly, overlap between the two approaches nominates cell types for targeted mechanistic follow-up, whereas non-overlap helps distinguish promoter-driven from vector-driven contributions. In addition, we wish to provide a comprehensive study regarding the identification of parenchymal cells edited by intravenous rAAV9 by examining 25 different tissues at multiple time points over a six-month period. Tissues were selected to span major organ systems and reported/putative rAAV9 targets and off-target sites, providing a broad atlas that reduces sampling bias inherent to focusing on a small subset of organs.

2. Materials and Methods

2.1. Animals

The procedures and care for the mice were carried out in compliance with protocols approved by the Boise State University Institutional Animal Care and Use Committee (approval: AC21-013). The mice were housed in the Boise State University rodent vivarium at a temperature of 25 °C, with a light/dark cycle of 12 h and with food provided ad libitum. The B6 Ai9 Cg-Gt(ROSA)26Sortm9(CAG-tdTomato)Hze reporter mice (stock number: 007909) [22] were obtained from Jackson Laboratory (Bar Harbor, ME, USA). The cell lineage tracing experiments were initiated in 4-to-5-month-old Ai9 mice by retro-orbital injection of rAAV9 carrying a CMV promoter-driven CRE recombinase (3 × 10^12^ vg/mouse; Virovek, Inc., Houston, TX, USA; Figure S7) [23]. Both male and female mice were used in this study, and there was no observed sex-linked variation in lineage tracing. The genotyping for transgenic mice was performed as previously described [22]. Healthy adult mice (4–5 months; both sexes) surviving to terminal harvest were included; pre-defined exclusions (non-study illness, dosing failure, gross processing failure) did not occur, and no data were excluded. Groups were defined only by the terminal time point (1.5, 3, 6 months), so no randomization was used. Tissues were labeled by ear tag ID only, and processing/analysis was performed by a student blinded to cohort/time point and not involved in injections.

2.2. Sample Processing and Sectioning

The mice used in the study were first anesthetized with isoflurane. Then, they underwent transcardial perfusion with a solution of heparin (20 U/mL) in phosphate buffer (PB) followed by 4% paraformaldehyde (PFA) in PB. The tissue was collected, placed in ice-cold 4% PFA/PB, and stored at 4 °C for 24 h. Afterward, the samples were placed in a solution of 30% sucrose/PB for an additional 72 h at 4 °C. The tissue was then blotted with a chemwipe and frozen quickly in OCT media on a block of dry ice. The samples were wrapped in aluminum foil, placed in freezer bags, and stored at −80 °C until they were ready to be sectioned. The tissue was equilibrated to −20 °C before being sectioned with a cryostat (Leica CM1950, Nussloch, Germany) at a thickness of 15 µm and then placed directly onto Superfrost Plus Gold microscope slides (Thermo Fisher, Waltham, MA, USA). The slides were left to dry overnight in a dark drawer. The following day, the slides were either processed for labeling or placed in a slide holder, sealed and stored at −80 °C.

2.3. Samples Processed for Hoechst Dye Labeling Only

The microscope slides were taken out of storage at −80 °C and left to reach room temperature (RT) in a dark drawer. Slides were then washed with PB and then treated with a PB solution containing Hoechst 33342 dye (Thermo Fisher; H3570; 1:20,000) for a duration of 10 min at RT. After that, one PB wash was performed, and the sections were left to dry for 30 min in a dark drawer. Then, coverslips were applied to the slides using Everbrite mounting media (Biotium, Fremont, CA, USA; 23003). The slides were left to dry for 24 h in the dark at RT, and then they were stored in a slide box at 4 °C until they were ready to be viewed.

2.4. Immunohistochemistry

The process of antigen retrieval by citrate treatment was carried out as previously described [24]. The tissue slides were taken out of storage at −80 °C and allowed to acclimate to RT in the dark for one hour. The slides were washed with PB and then placed in a container with citrate solution (Cell Signaling, Danvers, MA, USA; 14746S) that was diluted in distilled water and preheated in a microwave (on “HIGH” setting) for 5 min. The container and its contents were heated in the microwave (on “HIGH” setting) for an additional seven minutes and then covered and placed in an ice bath for one hour. Afterwards, the slides were washed three times with a Tris solution (not pH adjusted) with 0.05% Triton X100 and then with a Tris solution (pH 7.2) containing 0.05% Triton X100 (TBST). The slides were dried for 30 min, and then primary antibodies were applied in TBST supplemented with donkey serum at a 4% concentration. The primary antibodies used in the study were anti-mCherry (Biosource, King of Prussia, PA, USA; MBS448092; 1:100), anti-NeuN (Cell Signaling Technology; 24307S; 1:100), anti-insulin (Cell Signaling Technology; 3014S; 1:800), anti-glucagon (Santa Cruz Biotechnology, Dallas, TX, USA; sc-514592, 1:100), anti-calbindin (Cell Signaling Technology; 13176S; 1:100) and anti-MYH (Santa Cruz Biotechnology; sc-376157; 1:100). The slides were incubated overnight at room temperature in a humidifying slide chamber. The next day, the slides were washed with TBST and then incubated with the appropriate secondary antibodies that were diluted in TBST with donkey serum at a 4% concentration for 2 h at RT. The secondary antibodies used in the study were donkey anti-goat Alexa 594 (Thermo Fisher; A32758; 1:500), donkey anti-mouse Alexa 488 (Thermo Fisher; A-21202; 1:500) and donkey anti-rabbit Alexa 488 (Thermo Fisher; A-21206; 1:500). The slides were washed with TBST and then incubated with a Hoechst 33342 dye solution (1:20,000) for 10 min at RT. A final wash was done with PB, and the slides were dried for 1 h in the dark. Afterward, the slides were mounted with coverslips using Everbrite mounting media. The coverslips were allowed to dry for 24 h in the dark at RT before being imaged or stored at 4 °C.

2.5. Sample Imaging

Standard fluorescence microscopy images were captured using an EVOS M5000 Imaging System (Thermo Fisher) and viewed through a DAPI filter cube for Hoechst images and an RFP filter cube for tdTomato images. All confocal images were captured using a Leica Stellaris 5 Confocal System. Furthermore, a supplementary database containing microscopy images from all the analyzed tissues can be found on the Open Science Framework (https://osf.io/dpa29, accessed on 1 March 2026).

2.6. Image Processing and Cell Quantification

Automated quantifications of tdTomato-expressing cells on prepared tissue slides were imaged using an EVOS M5000 system (Fisher Scientific, Carlsbad, CA, USA). Positive tdTomato fluorescing cells and Hoechst-stained nuclei were counted in an automated fashion using ImageJ (version 1.54i, NIH in Bethesda, MD, USA). Briefly, blue-only and red-only images were converted to 8-bit images in ImageJ, and the threshold was adjusted to reduce the background. The watershed tool was used to segment adjacent particles. Finally, the “analyze particles” function was utilized to set size restrictions for cells according to the size of a representative cell for each tissue type. These one-time settings were used for all images of a given tissue across all cohorts to allow for direct comparisons. The percent of tdTomato-positive cells was determined by positive tdTomato cells/total Hoechst nuclei per field, as tabulated by ImageJ in this automated workflow.

Confocal microscopy with a Leica Stellaris 5 Confocal System was used to determine tdTomato co-localization with cell markers following immunohistochemistry. Image stacks of 20 um were taken per field with at least three fields per mouse (three mice per time point). Orthogonal analysis was utilized to verify co-localization of the tdTomato signal with the indicated cell marker signal. Cells were tabulated according to co-localization status, and statistical analysis was performed.

2.7. Statistical Analysis

Statistical analysis was performed using GraphPad Prism 6 (GraphPad Software, San Diego, CA, USA), where datasets consisted of three replicates (three separate tissue sections) averaged for each of the three mice per time point. No a priori power calculation was performed (exploratory mapping study); n = 3 mice/time point was chosen for feasible biological replication across targets and time points. Given the exploratory mapping design and small n, inferential statistics are presented descriptively, and effect estimates should be interpreted cautiously; mean ± SEM is shown primarily to visualize variability across mice. All cell count and fluorescence intensity data were evaluated using a two-way ANOVA followed by a post hoc Tukey’s test.

3. Results

3.1. Intravenous rAAV9 Model of Vascular Compartment Cell Lineage Tracing in Adult Mice

Previous work [21] has shown that numerous and diverse cell types are labeled following tracing from cells expressing the endothelial marker vascular endothelial cadherin (VEcad). The same study also observed diverse tracing by examining a more promiscuous endothelial marker, nestin, as it is expressed by subsets of endothelial cells and endothelial progenitor cells [25,26,27,28]. Some cell types were labeled by both VEcad- and nestin-driven cell lineage tracing, suggesting a possible endothelial origin in adult mice. However, these studies relied on the expression of a promoter to initiate labeling following tamoxifen treatment, which could lead to off-target labeling events. Therefore, we utilized a distinct labeling strategy for this current study that would target all cells of the vasculature, including endothelial cells, with the goal of identifying traced cells found in both studies. It is important to note that rAAV can undergo transcytosis into the parenchyma; therefore, some parenchymal labeling could arise from direct viral passage/transduction rather than transfer from vascular cells, which limits mechanistic inference from this dataset alone. This work would also more broadly shed light on the dynamics and cell type targeting of rAAV9-directed intravenous gene therapy. We performed cell tracing by intravenously delivering rAAV serotype 9 (rAAV9) carrying a CMV promoter-driven CRE recombinase (3 × 10^12^ vg/mouse), a dose well-established for transgene delivery [23], to 4-to-5-month-old Ai9 mice containing a flox-STOP-tdTomato reporter (Figure 1). After intravenous rAAV9 transduction in adult transgenic mice, vascular compartment cells experience CMV promoter-driven CRE expression, which results in the excision of a pLox cassette containing a translation stop codon within a separate tdTomato fluorescent reporter gene. Removal of this stop sequence permits constitutive tdTomato expression via the Rosa26 CAG reporter. Importantly, because Cre-mediated recombination is permanent, tdTomato positivity reflects historical exposure to sufficient Cre activity and does not imply persistent episomal AAV genomes or ongoing CMV-driven expression; conversely, lack of tdTomato does not exclude transient AAV transduction that failed to reach a recombination threshold. It is important to note that the CAG promoter, while widely expressed, are not ubiquitously expressed nor consistently expressed (can be silenced) so an absence of labeling could represent a false negative result. Mice were then sacrificed at the same time intervals post-injection/labeling (1.5, 3, and 6 months) as in our previous study [21] for direct comparison. These time points were selected to capture (i) an early post-delivery interval after initial transduction and stabilization of reporter expression (1.5 months), (ii) an intermediate interval (3 months), and (iii) longer-term persistence/durability of labeling (6 months).

3.2. Cell Tracing in the Periphery by Intravenous rAAV9

A major goal in this rAAV9-mediated tracing study was to determine whether the novel labeling events observed in the VEcad-expression cell lineage tracing model from our previous study could have arisen from a vascular source consistent with endothelial cells. We observed tdTomato fluorescent reporter expression at 1.5, 3 and 6 months following rAAV9-CRE administration in numerous cell types and tissues in the periphery (Figure 2). Interestingly, a subset of these labeled cells was also identified in our previous study [21] that utilized VEcad tracing. These included skeletal myocytes, pancreatic islet cells, pancreatic acinar and small intestinal epithelial cells. We verified the identity of these cells by co-labeling with cell markers using confocal microscopy (Figure 3A). Tracing to these cells was also quantified for each post-injection time point by co-labeling with a marker and performing orthogonal analysis following confocal microscopy imaging (Figure 3B–D and Figures S1–S3). Skeletal myocytes, pancreatic beta cells, pancreatic acinar cells and duodenal epithelial cells all exhibited an increase in traced cells over time.

3.3. Central Nervous System Cell Lineage Tracing by Intravenous rAAV9

The central nervous systems of the Ai9 transgenic mice were also examined following intravenous rAAV9-mediated tracing, and we observed tracing events to hippocampal neurons and Purkinje cells (Figure 4). Interestingly, we found no overlapping events when compared to our previous VEcad-expression tracing methods. This suggested that a blood compartment connection not of endothelial cell origin was a likely origin of tracing. However, due to the disease relevance of hippocampal neurons and Purkinje cells, we confirmed tracing by co-localization with cell marker antibodies and confocal microscopy (Figure 5A). Tracing and cell marker colocalization were quantified by orthogonal analysis (Figure 5B,C, Figures S4 and S5). No change was observed for hippocampal neurons, while Purkinje cells showed an increase over time. It should be noted that extremely rare tracing events to unidentified neurons (NeuN+) were observed at a rate of 0–10 per brain cross-section (Figure S6).

Quantification for all cell lineage tracing events by tissue, regardless of cell type, using conventional fluorescence microscopy, as represented in Figure 2 and Figure 4 are shown (Figure 6). Note that all examined cardiac myocytes showed tdTomato fluorescence (compared with non-injected Ai9 mice) following rAAV administration (Figure 2). Therefore, fluorescence/area for cardiac tissue was quantified across all time points. A simplified summary of all tracing events that included negative results is displayed in Table 1.

4. Discussion

Previous studies have indicated that intravenous recombinant rAAV delivery results in highly efficient transgene delivery to parenchymal cells, such as cardiac myocytes and skeletal myocytes, in mammals [8,29]. However, the genetic editing mechanism of these parenchymal cell types for this promising route of clinical gene therapy is unproven, and existing paradigms (e.g., transcytosis by endothelial cells) do not comfortably explain this phenomenon, as alternative non-viral models suggest similar cell lineage tracing results [21]. Our prior work [21] raised the possibility that vascular endothelial cells might contribute to some downstream labeling events, while also highlighting interpretative constraints of promoter-driven tracing. That study found VEcad-promoter-driven tracing to several cell types in adult mice, prompting consideration of several non-mutually exclusive possibilities for downstream labeling after vascular exposure, including EV-associated transfer and viral transcytosis; endothelial cell fate change remains speculative and requires direct evidence. Note that our previous non-viral study [21] results indicate that transcytosis of viral particles is likely not a contributor for the cell types identified in that investigation, although it cannot be entirely ruled out as a coincidental effect in this rAAV9 study due to evidence of AAV transcytosis in other systems [18,19,20]. We sought to pursue the possibility of endothelial cell origins of tracing further by initiating cell tracing in the vascular compartment. Cells that exhibit tracing from the VEcad promoter [21] and intravenous rAAV9 would be prime candidates for endothelial cell connections. Also, other studies have largely focused on an individual rAAV serotype (of which there are hundreds of sub-types) for a particular target cell type or tissue in a specific, often disease-relevant context, thereby obscuring broader interpretations of results [30]. To overcome these limitations, we expanded the scope of our analysis to 25 distinct tissues and cell types and utilized a popular rAAV serotype, rAAV9, reported to readily transduce a broad array of cell types that include endothelial cells [30,31]. The same transgenic reporter mouse line was utilized as in our previous study [21], along with matching time points, which extended to 6 months, and tissue assessment parameters for direct comparison. Thus, this study adds to our understanding of possible endothelial cell contributions to gene therapy while also adding a sizeable dataset for reference by future studies into intravenous rAAV9.

Examination of tissue following intravenous rAAV9 administration revealed numerous diverse cell lineage tracing events. High transduction by rAAV9 for liver (hepatocytes) and heart (cardiac myocytes) is well-known and was also observed in our system [32]. We also witnessed rAAV9 tracing to murine pulmonary epithelium, as others have reported [33]. Our study concurs with previous investigations demonstrating transduction-tracing to pancreatic beta cells [34] and acinar cells [35] by intravenous rAAV9 in mice. These two pancreatic cells were also identified in our prior VEcad tracing study [21], which hints at a potential endothelial cell relationship. Previous reports indicate myenteric plexus labeling following systemic rAAV9 injection in neonatal, juvenile and adult mice when examined two-weeks post-injection [36,37]. We also observed tracing to the gastrointestinal myenteric plexus (Figure 2). Interestingly, we also found that gastric epithelial cells, duodenal epithelial cells, ileal epithelial cells and colonic epithelial cells exhibited rAAV9 tracing. Of these gastrointestinal-traced cells, duodenal and ileal epithelial cells were also observed in our previous VEcad tracing study, making them particularly relevant to possible endothelial cell connections. It has been reported that systemic injection of rAAV9 results in restricted labeling to the glomerulus (endothelial cells), and we observe this as well [38]. The slit filter restriction of ~50 kDa in the glomerulus likely prevents viruses from entering the filtrate/urine and directly transducing cells lining the nephron tubules. Moreover, labeled parenchymal cells were not observed, suggesting no transgene emission from the vascular compartment in this tissue. Like the liver, the spleen possesses fenestrated and discontinuous capillaries that allow for deep penetration of intravenous viral particles into the matrix, and this is supported by studies showing high levels of rAAV9 viral particle retention there [39]. However, unlike hepatocytes, we found only endothelial cells labeled by rAAV9 in the spleen (Figure 2). This implies that splenic B cells and T cells are resistant to rAAV9 infection or that these cells do not express the Ai9 tdTomato reporter gene following activation—a known limitation when assaying for reporter expression. Indeed, we cannot find evidence in the literature for lymphocyte/splenocyte transduction by rAAV9, suggesting that our findings are consistent with the current knowledge. Interestingly, this finding might bear relevance to the Purkinje cell tracing discussed in the section below.

Analysis of the CNS following systemic rAAV9 administration revealed tracing to hippocampal neurons, Purkinje cells and astrocytes. None of these groups exhibited VEcad-mediated labeling in our prior study [21], which does not support an endothelial-lineage explanation for these CNS labeling patterns in the present dataset. However, intravenous tracing to these disease-relevant cell types prompted a closer examination since systemic rAAV9 delivery is a promising approach to targeting these cells for genetic therapy. It has been reported that intravenous rAAV9 traces to hippocampal and Purkinje cells, as we also found [39]. These prior studies primarily measured transient transgene expression from episomal rAAV genomes (e.g., eGFP), whereas here we assay permanent Cre-mediated genomic reporter activation in Ai9 mice; the two readouts are related but not equivalent. We also witnessed tracing events to astrocytes, primarily enriched in the midbrain and hippocampal regions (Figure 4), which has been previously reported with intravenous rAAV9 [39,40]. Interestingly, there is reported evidence that the tracing to neurons versus astrocytes is dramatically affected by the age of the mice (neonate vs. adult) at the time of injection [40]. This might indicate different mechanisms of tracing for early development versus adult tissue and is consistent with the idea that adult tissues have their own distinct processes for homeostasis and cell replacement. In fact, this has been an emerging field of study in and of itself over the past few years, largely as the result of single-cell RNA sequencing [41,42]. This idea is at the center of interest in ascertaining whether endothelial cells serve as a progenitor pool via transdifferentiation in adult mice. Of course, if this is a natural phenomenon, it is important to identify when this process does not likely occur as our data suggests for the CNS. We also did not detect rAAV9 tracing to microglia in the CNS, which is consistent with other studies [23]. Lastly, we did not observe significant neuronal tracing in the olfactory bulb except in extremely rare events (Figure S6).

The precise routes by which rAAV delivers transgenic cargo in vivo to the parenchyma are unknown. Considerable resources are being directed toward identifying natural AAV variants or creating new recombinant ones to tailor the tropism for rAAV vectors. However, there are large assumptions being made about the mechanism of transduction from systemic delivery to parenchymal cells. A major underlying assumption is that the viral capsid is transcytosed into the parenchymal space, and cell tropism is then determined by viral capsid protein interactions with host cell glycans and proteoglycans in this parenchymal environment. Compelling evidence has long uncovered numerous instances in which this is not the case. For example, Purkinje cells have been reported to be labeled via circulating lymphocyte fusion events [43]. Thus, growing reports of Purkinje cell transduction following intravenous injection of rAAV are greatly confounded [44,45]. We also observed Purkinje cell tracing in our present study using rAAV9, which is consistent with a circulating cell origin. Whether these traced cells are fully functional and long-lived is a matter of contested debate [46], but their mere presence requires a more empirical examination of presumed intravenous rAAV transduction mechanisms. These fusion events are akin to transdifferentiation, in which transduction of one cell type leads to the subsequent labeling of another cell type through the inheritance of an altered or additional nuclear genome (e.g., CRE-lox reporter expression). A major interest of our work is to uncover whether endothelial cell transdifferentiation to embryonically distinct cells occurs as a natural homeostatic mechanism. Likewise, we are interested in whether endothelial cell EVs are transmitting gene-editing machinery from the vasculature to the parenchyma following intravenous rAAV9 injection. Here, we have identified five cell types that are consistent with these ideas when compared to our previous VEcad tracing study. It is important to note that our rAAV9 tracing findings might also result from rAAV transcytosis to the parenchyma. Future alternative tracing approaches will be needed to rule out any transcytosis contributions. One such approach could utilize a nuclear genome labeling approach (unexpressed barcode) in the vasculature since rAAV can be transmitted by cytosol (via EVs) but only rarely integrates into the host genome.

It has been reported that rAAV can associate with EVs and constitute a transmissible vector [47], and this observation has even led to a new approach for rAAV engineering, termed exo-AAV [48]. This revelation opens the possibility of secondary transmission of rAAV. One initial infection event inside the vasculature results in transduction of an unrelated parenchymal cell type via emitted EVs. In such a case, the rAAV serotype might be more pertinent to the cell type of initial transduction than to the eventual parenchymal target. More recent work has further characterized EV-associated AAV preparations and reported enhanced transduction in barrier contexts, supporting the plausibility that EV association can modulate AAV cellular entry and downstream spread [49,50]. In parallel, improved understanding of AAV attachment/entry and intracellular trafficking emphasizes that endocytic routing and post-entry processing are rate-limiting and cell-type dependent determinants of transduction outcomes [51]. The finding that VEcad tracing [21] has some commonality with intravenous rAAV9 also raises the possibility that some of these tracing events are independent of the rAAV9 vehicle and might be endothelial cell in nature. Taken together, these less-considered transduction routes show the importance of utilizing intact animal models to evaluate rAAV biology versus cell culture systems and might explain the diverse cell types labeled in our study (Table 1) and others that differ considerably from the cell culture screens reported for rAAV9 tropism [52].

In this study, we mapped long-term, tissue-wide reporter activation after a single intravenous dose of rAAV9-CMV-Cre in adult Ai9 mice. Because this approach cannot distinguish direct parenchymal transduction from indirect vascular-to-parenchymal transfer, overlap with VEcad-promoter labeling should be treated as a prioritization tool for mechanistic follow-up rather than evidence of endothelial lineage. Accordingly, skeletal myocytes, pancreatic beta/acinar cells, and duodenal/ileal epithelia emerge as pragmatic targets for future experiments that explicitly test transcytosis versus secondary transfer routes. More broadly, the atlas provides a durable reference for interpreting systemic rAAV9 labeling over multi-month intervals.

Limitations of This Study

This study used a broadly active CMV promoter to drive Cre recombinase as part of an atlas-style assessment of intravenous rAAV9 biodistribution and downstream reporter labeling rather than a tissue-restricted therapeutic design. Accordingly, the present dataset does not test whether tissue-specific promoters (or other targeting strategies) would mitigate off-target labeling, and this remains an important direction for future work. In addition, the atlas-style design used a small cohort size (n = 3 mice per time point) and was not powered to detect subtle differences across time or tissues; accordingly, these data should be interpreted as a descriptive resource and hypothesis generator rather than a definitive quantitative comparison.

Safety is a main limiting factor in systemic rAAV-mediated gene delivery. Despite the use of tissue-specific promoters, there is still a possibility of “off-target” effects due to vector uptake by tissues outside of the target, and capsid tropism, promoter leakiness, and biodistribution variability among individuals. Dose-related toxicities have been reported with the intravenous administration of multiple rAAV serotypes, and host responses to the vector or transgene, which can cause inflammation in target tissues [53,54,55].

Direct and intravenous delivery of several rAAV serotypes has been shown to be toxic to mammalian neurons and other cell types when used at common experimental doses [53,54]. However, this would likely result in the under-reporting of traced cells and likely not affect the reliability of the observed traced cells in this study. Our study utilizes a transgene reporter that was inserted into a single known locus (Rosa26) in the host Ai9 mice [22]. However, whether a cell type expresses this locus after editing by the CRE delivered via intravenous rAAV9 is not certain. Not all cells expressed the reporter in this Ai9 mouse strain. This is evidenced by negative results from CRE line crosses expected to express a particular cell type. Thus, this phenomenon would likely result in an under-reporting of tracing events. Nevertheless, there is a key advantage of the Ai9 system: a brief Cre pulse permanently recombines the reporter locus, enabling durable readout of historical Cre activity. This durability should not be interpreted as durable AAV genome persistence or durable transgene expression, which can wane over time. It is well known that transgene expression by AAV is diminished over time due to a number of factors [55]. In addition, CMV promoter activity can be attenuated or silenced in a cell type- and context-dependent manner, which could reduce Cre expression and produce false-negative labeling in some tissues. This was of particular concern for our three- and six-month time points, which prompted us to employ the Ai9 mouse tracing strategy used in this study.

5. Conclusions

By utilizing a vascularly initiated CRE reporter strategy within adult Ai9 mice, we hope that a time-resolved assessment of downstream parenchymal labeling post-administration of a single dose of rAAV9 via the intravenous route will allow us to determine the tissues that exhibit persistent reporter gene activity post-administration. These studies should serve as a basis for future mechanistic studies that hope to disentangle the possible mechanisms of downstream labeling post-administration, such as transcytosis and extracellular vesicle-mediated transcytosis, and transdifferentiation.