Selective targeting of striatal parvalbumin-expressing interneurons for transgene delivery

PVCre mice combined with AAV-FLEX vectors allowed efficient and specific targeting of PV+ interneurons in the striatum. However, diffusion of viral particles to the globus pallidus caused massive transduction of PV+ projection neurons and subsequent anterograde transport of the transgene product to the subthalamic nucleus and the substantia nigra pars reticulata. Different AAV serotypes (1 and 9) and promoters (CBA and human synapsin) were evaluated. The combination of AAV1, a moderate expression level (human synapsin promoter) and a precise adjustement of the stereotaxic coordinates in the anterior and dorsolateral part of the striatum were necessary to avoid transduction of PV+ GP projection neurons. Even in the absence of direct transduction due to diffusion of viral particles, GP PV+ projection neurons could be retrogradely transduced via their terminals present in the dorsal striatum. However, in the absence of diffusion, GP-Str PV+ projection neurons were poorly (AAV9) or not (AAV1) transduced suggesting that retrograde transduction did not significantly impair the selective targeting of striatal PV+ neurons. Finally, a prominent (about 50%) toxicity of the Cre recombinase toward the striatal PV+ interneurons was evidenced suggesting that functional effects of AAV-mediated transgene expression in PV+ striatal interneurons in PVCre mice should be analyzed with caution.


Introduction
AAV vectors have gained increasing interest for gene therapy of neurological diseases (Kantor et al., 2014;Hocquemiller et al., 2016) as well as for functional studies in neuroscience research (Bedbrook et al., 2018). Targeting specific neuronal subpopulations is a rapidly growing field allowing the manipulation of neuronal circuits using optogenetics (Hunnicutt et al., 2016) or chemogenetics (Woloszynowska-Fraser et al., 2017), the modeling of disease using human disease-causing transgenes (Grames et al., 2018) as well as next-generation targeted gene therapy (Chtarto et al., 2013;Dalkara et al., 2013;Vormstein-Schneider et al., 2020). Thanks to their ability to be axonally transported (Castle et al., 2014), AAV vectors are also useful for anatomical tracing (Zingg et al., 2017).
The development of vectors selectively targeting neuronal subpopulations is coming of age. Some examples are: GluA4-AAV a capsid variant designed to attach to glutamate receptor 4 (GluA4), selectively expressed by parvalbumin-positive (PV+) interneurons (Geiger et al., 1995), combined with the SFFV (spleen focus-forming virus) promoter (Hartmann et al., 2019); AAV-mDlx a vector targeting interneurons using mDlx transcriptional regulatory elements combined with AAV9 (Dimidschstein et al., 2016) or AAV5 (Lee et al., 2014) capsids; AAV-TH which is targeting dopaminergic neurons thanks to the use of a fragment of the tyrosine hydroxylase promoter (Stauffer et al., 2016) and AAV PHP.eB with hybrid promoters containing specific enhancers targeting PV-and VIP-interneurons (Vormstein-Schneider et al., 2020).
With the advent of single-cell RNA sequencing (Munoz-Manchado et al., 2018;Gokce et al., 2016) and methods for mapping chromatin accessibility (Buenrostro et al., 2015), new synthetic promoters will eventually be identified, which will further refine these molecular tools (Juttner et al., 2019).
Targeting specific neuronal subpopulations can also be achieved using a combination of Cre-driver mice (Hippenmeyer et al., 2005) or rats  expressing the Cre recombinase under the control of a cell-type specific gene and an AAV vector harboring an inverted ORF flanked by 2 pairs of Cre recognition sites positioned so that expression occurs only when the Cre protein is present (Saunders et al., 2012;Saunders and Sabatini, 2015).
In the striatum, projection neurons, also called medium-sized spiny neurons, relay motor output, expressing D1R-or D2R-type of dopamine receptors which, in response to dopamine respectively activate (D1R) or inhibit (D2R) efferent structures of the motor loop (Surmeier et al., 2007). The striatum also contains several classes of interneurons among which the (PV+) fast spiking and the cholinergic neurons which are thought to coordinate the activity of the projection neurons (Gritton et al., 2019).
In the present study, we have focused on the targeting of PV + interneurons of the dorso-lateral striatum which are key to the control of the sensorimotor striatum (Lee et al., 2017) using PV Cre driver mice (Hippenmeyer et al., 2005).
Limitations of the AAV-FLEX/Cre driver mice targeting system have been previously described: i) off-target expression in cells not expressing Cre (Fischer et al., 2019), ii) expression of Cre in only a part of the targeted neuronal population (Saunders et al., 2016), iii) Cre recombinase-induced cellular abnormalities (He et al., 2014).
We show here that, despite the efficient and specific cellular targeting offered by the Cre-lox system, the AAV delivery parameters have to be precisely adjusted to selectively target the striatum with exclusion of the adjacent globus pallidus containing a high density of PV + projection neurons. Furthermore, PV + striatal interneurons were decreased by the Cre-recombinase. These data suggest that functional effects of AAV-FLEX transgene expression in PV + striatal interneurons in PV Cre mice should be analyzed with caution.
Eleven weeks-old homozygous PV Cre mice and wild-type (WT) C57Bl6-Ola-Hsd mice (En Vigo) of both sexes were used for all experiments.

Stereotaxic injections
Adult mice (11 weeks-old) were used for unilateral intrastriatal injections. Briefly, the animals were anesthetized with a mixture of ketamine (100 mg/kg, Ketasol, Graeub AG) and xylasine (10 mg/kg, Rompun, Bayer). Injections were made according to coordinates defined by "The Mouse Brain in stereotaxic coordinates, 3rd edition, Franklin, K. B.J. and Paxinos, G. AcademicPress, 2007" using a Kopf stereotaxic apparatus (David Kopf, Tujunga, California). Viral particles diluted in 1 μl of D-PBS (Biowhittaker, Lonza) were infused in the striatum, using a 34 G needle at different coordinates (see Table 1). After injection, the needle was left in place for 5 min in order to allow diffusion of the viral suspension in the parenchyma. The needle was then slowly removed. Animals were maintained in a 12:12 h light-dark cycle with free access to food and water.
Experimental procedures were approved by the "Affaires vétérinaires" of the Canton de Vaud" (Authorization n • VD3400).

Brain collection and immunohistochemistry
Two weeks after viral injection, mice were euthanized with an overdose of pentobarbital (30 mg/kg in 0.9 % NaCl). A 4 % paraformaldehyde (PFA) solution at pH 7.4 in phosphate buffer saline (PBS, Bichsel AG) was freshly prepared before use. The mice were transcardially perfused consecutively with a PBS solution at pH 7.4 and with the ice-cold 4 % PFA solution. Brains were collected and post-fixed in 4 % PFA overnight at 4 • C. Consecutive incubations of 24 h in 20 % and 30 % sucrose solutions were performed to cryoprotect the brains which were then slowly frozen by consecutive immersions in 2-methyl-butane at − 10 • C and− 20 • C and finally stored at − 80 • C. A cryostat (Leica Biosystems, CM1850) was used to collect 25μm-thick coronal sections which were stored in an anti-freeze solution (glycerol 25 %, ethylene glycol 30 % and Na-phosphate buffer 50 mM) at − 20 • C. The following antibodies were used to stain PV + cells: guinea pig anti-PV (1:1000, cat. #195004, Synaptic Systems, Göttingen, Germany), biotinylated goat anti-guinea pig (1:200, cat. #BA-7000, Vector Laboratories, Burlingame. USA) and Cy3-conjugated streptavidin (1:300, cat. #016-160-084, Jackson ImmunoResearch Laboratories, West Grove -USA). Free floating sections were stained as follows. Sections were washed 3 times for 10 min in Tris-buffered saline (TBS, 10 mM Tris pH 7.6 and 0.9 % NaCl) at room temperature (RT). Then, they were incubated 1 h at RT in a blocking solution composed of 5 % bovine serum albumin (BSA) in THST buffer (50 mM Tris pH 7.6, 0.5 M NaCl and Triton X-RT100 0.5 %). Afterwards, they were incubated overnight at 4 • C with the primary antibody in a THST solution containing 1 % BSA. The second day, sections were first washed 3 times for 10 min in TBS at RT and then incubated for 1 h at RT with the secondary antibody in THST. Sections were then washed 3 times for 10 min in TBS and incubated in the dark with the streptavidin conjugate for 1 h at RT in THST. Finally, the sections were washed 3 times in PBS for 10 min at RT. The sections were mounted on microscope slides and covered with Vectashield mounting medium (Vector Laboratories, Burlingame, USA).

Image acquisition and quantifications
Whole slide images were taken with a Zeiss Axioscan Z.1 slidescanner (Carl Zeiss Microscopy, Germany) using a Plan-Apochromat 10x/0.45 or a Plan-Apochromat 20x/0.8 objective. All the images were taken with an Orca-Flash 4.0 V2 digital CMOS camera. 16 bits images were obtained. Cy3 was excited at 553 nm with a 555/30 nm LED at 50 % power. A beam splitter at 568 nm was used. The detection range was 578− 640 nm. eGFP was excited at 493 nm with a 469/38 nm LED at 20 % of power. A beam splitter at 498 was used. Detection range was 507− 546 nm. Confocal images were taken with a confocal microscope Zeiss LSM 800 (Carl Zeiss Microscopy, Germany) equipped with a 3x GaAsp detector. All images were collected with a Plan Apochromat 20×/0.8 DIC II objective with a pinhole set at 36 μm. 8 bits resolution images have been obtained by bidirectional scanning and 4x averaging. Cy3 was excited with a 561 nm laser at 0.03 % of power. 739 V of master gain was applied and the detection range was 566− 628 nm. eGFP was excited with a 488 nm laser at 0.03 % of power. 750 V of master gain was applied and the detection range was 410− 546 nm. Controls with singlefluorescence were performed. No interference between red and green fluorescence was observed.
Images were processed using Zen Blue 2.3 (Carl Zeiss Microscopy, Germany) and ImageJ/Fiji. Z-stacks were acquired on a 25 μm thickness.
Shown confocal images in Figs. 1, 2, 5 and 6, are 2D maximal intensity projections of 35 images. Cells were manually counted on 2D maximal intensity projections of confocal images transformed into the tiff format. Cells were counted by a blind observer using the Multi-point tool in ImageJ. For evaluating the specificity, GFP + , PV + , and double-labeled cells were counted. For each animal, five brain sections were selected, and three pictures were taken for each section. To evaluate the efficacy in the dorsal striatum, the striatum was divided in two by a horizontal line, and GFP + , PV + , and double-labeled cells were counted in the upper part.
For the quantification of native GFP fluorescence intensity of individual cells (Fig. 6), confocal images were used and the cells were delineated using the "free-hand" selection tool of the ImageJ software. The mean fluorescence intensity was recorded for each individual cell. Area with a similar surface in non-transduced area of the GP or striatum were measured and the obtained values substracted.-In order to evaluate Cre recombinase effect on PV + cells, WT and PV Cre mice were compared. For each AAV-FLEX virus, three WT and three PV Cre successfully injected animals were available. The number of PV + cells in the non-injected hemisphere was counted on a section from the same region for each animal and divided by the surface in mm 2 . The images were transformed in. jpeg format at a resolution of 300dpi and countings were performed by a blind observer using the Multi-point tool in ImageJ.
To determine viral toxicity the number of PV + cells of the injected hemisphere was compared to the corresponding area in the non-injected hemisphere. In order to count cells in the whole striatum, images were acquired with the slide scanner. The images were transformed in. jpeg at a resolution of 300dpi format and cells were counted by a blind observer using the Multi-point tool in ImageJ.

Statistical analysis
Data analyses and the creation of graphs were performed using GraphPad Prism 8 software (San Diego, CA) for Windows. Comparisons were performed by One-way ANOVA followed by Tukey's multiple comparisons test (Table 2), two-way ANOVA followed by Sidak post hoc test (for more than 2 groups) or by Student t-test (2 groups). Results were expressed as mean ± SD and statistical significance was established for a p value ≤0.05.

Efficiency and cellular specificity of transgene expression in striatal PV + interneurons mediated by AAV2/9-FLEX vectors in PV Cre mice
AAV-FLEX vectors and PV Cre mice were used to target transgene expression into striatal PV + interneurons. Since PV expression was weaker in the striatum than in other brain regions (for example in the cerebral cortex; see Suppl. Fig. 1), we used AAV-FLEX vectors (Oh et al., 2014) with strong promoters; the non-specific CMV/chicken β-actin (CBA) promoter (Burger et al., 2004;Klein et al., 2002) or the neuron-specific human synapsin promoter (hsyn) (Shevtsova et al., 2005) combined with the "Woodchuck hepatitis virus post-transcriptional regulatory element" (WPRE) which enhances mRNA stability and protein synthesis (Klein et al., 2002;Oh et al., 2014). In order to further maximize transduction efficiency, the vectors were encapsidated into AAV9 serotype (Cearley and Wolfe, 2006;Klein et al., 2008).
The vectors (7.9 × 10 9 vg in 1 μl) were injected in the striatum of homozygous PV Cre mice using various coordinates (see Table 1). As a Table 1 Summary of the coordinates sets and of the number of animals used for each AAV-FLEX virus.

Virus
Coordinates ( negative control, the vectors were also injected in WT C57/Bl6 mice. GFP + cells were observed in the striatum with a distribution corresponding to the expected pattern of PV + interneurons (Fig. 1A & E). GFP + fibers were more strongly labeled with the CBA promoter ( Fig. 1B & D) as compared with the hsyn promoter ( Fig. 1F & H), suggesting, that, as expected, the hybrid (cellular/viral) CBA promoter drives a higher transgene expression level than the cellular hsyn promoter.
In order to confirm the cell-type specificity of transgene expression, brain sections were labeled with anti-PV antibodies.
The efficiency of PV + interneurons transduction, evaluated as the number of double-labeled PV + /GFP + cells (Fig. 1D & H), relative to the total number of PV + cells (Fig. 1C & G) was similar (>90 %) for both promoters in AAV2/9 vectors (n = 5 for hsyn and n = 3 for CBA) ( Table 2). The specificity of the targeting, evaluated as the number of double-labeled PV + /GFP + cells relative to the total number of GFP + cells (Fig. 1B & F), was approx. 85-90 % for both promoters (Table 2). Few GFP + cells which did not show a detectable PV staining were observed (Fig. 1, arrow), suggesting that a low level of non-specific expression could have occurred. However, injection into WT C57/Bl6 mice did not reveal any GFP + cell (data not shown), suggesting that, as expected GFP expression was dependent on the presence of the Cre recombinase. The GFP + cells which were not labeled by anti-PV antibodies could be due to  a low, undetectable PV expression in these neurons or to a potential toxicity of the Cre recombinase (see below).

Transduction of contiguous and axonally-connected brain regions
Examination of the whole brains revealed GFP + cells in the globus pallidus (GP) (Fig. 2E) and in the cerebral cortex (CTX) ( Fig. 2A) in a large proportion of the animals. These data suggest that viral particles diffused to these neighboring regions which contain a higher density of PV + neurons as compared to the striatum (Saunders et al., 2016).
The localized cortical transduction was probably due to a reflux of the viral suspension along the needle and could be avoid using antireflux needles (Vazquez et al., 2012;Casanova et al., 2014;Lueshen et al., 2017).
The majority of GFP + cells also expressed PV in the cerebral cortex ( Fig. 2B-D) as well as in the GP (Fig. 2F-H).
Preferential transduction of the GP after injection of AAV vectors in the striatum has been previously reported (Tenenbaum et al., 2000) and was suggested to reflect a preferential diffusion of the viral particles along the vessels driven by the perivascular pump (Hadaczek et al., 2006). Several stereotaxic coordinates were tested in order to reduce diffusion to extra-striatal areas (see Table 1). The anteroposterior coordinate varied between +1.0 and +1.2, the dorsoventral coordinate varied between -2.75 and -3.0 and mediolateral coordinate was set at -1.8. For both promoters, the majority of the mice harbored a widespread transduction of the GP (see Table 1). With none of the coordinates was the GP transduction avoid in all mice.

Anterograde transport of GFP from GP PV + neurons projecting to the SNr and STN
GP contains 40 % of PV + neurons which project to the subthalamic nucleus (STN), to the substantia nigra pars reticulata (SNr) or to the striatum (Saunders et al., 2016).
Consistently, in animals with GP transduction, GFP + fibers were evidenced in the STN (Fig. 3A & B) and in the SNr (Fig. 3C & D).
In the striatum, it was not possible to distinguish GFP + fibers originating from striatal interneurons and pallido-striatal projection neurons. A typical pattern of pallido-striatal fiber tracts was evidenced in WT mice by staining with an anti-PV antibody ( Fig. 4A & B). In PV Cre mice in which striatal AAV2/9 injection resulted in transduction of GP PV + neurons, fiber tracts with the same pattern were GFP + (Fig. 4C & D), suggesting that, as expected they originate from the pallido-striatal projection neurons.

Efficiency and cellular specificity of transgene expression in striatal PV + interneurons mediated by AAV2/1-FLEX vectors in PV Cre mice
Since AAV9 vectors diffusion to the GP and to the CTX hindered the selective targeting of the striatum, we used the AAV1 serotype which has previously been shown to allow region-targeted striatal transduction in Table 2 Percentage of PV/GFP double-labeled cells among total GFP + cells (specificity) or total PV + cells (efficiency) in the dorsal striatum.
Therefore, the hsyn promoter was selected for further experiments. Several stereotaxic coordinates were evaluated to avoid GP transduction. Anteroposterior coordinates varied between +1.0 and +1.2 and dorsoventral coordinates between -2.75 and -3.0. The mediolateral coordinates were set at -1.8. A total of 5 mice were tested. GP transduction could be avoided only with the most anterior and dorsal coordinates (AP=+1.2; ML=− 1.8; DV=-2.75) (n = 3) (Fig. 5A).
The efficiency of PV + interneurons transduction, evaluated as the number of double-labeled PV + /GFP + cells (Fig. 5D), relative to the total number of PV + cells (Fig. 5C) was approx. 93 %) ( Table 2). The specificity of the targeting, evaluated as the number of double-labeled PV + / GFP + cells relative to the total number of GFP + cells (Fig. 5B), was higher than approx. 89 % ( Table 2). As for AAV9, injection into wild-type mice did not result in GFP + cells.
In order to confirm on a larger number of animals that the established stereotaxic coordinates (AP=+1.2; ML=− 1.8; DV=-2.75) allow to avoid AAV1 viral particles diffusion to the GP, 4 additional mice, taken from a different littermate of PV Cre (+/+) mice, were injected with AAV2/1-hsyn-FLEX-eGFP. However, in this littermate, the loss of PV + cells in the striatum was aggravated and only few PV + and GFP + cells were observed, precluding a relevant statistical analysis (Suppl. Fig. 2A). However, despite an apparently normal amount of PV + cells in the GP, no GFP + cells were observed (Suppl. Fig. 2B).

Retrograde transduction of pallido-striatal neurons by AAV-FLEX vectors?
Since vectors were injected in the dorsolateral striatum in which pallido-striatal PV + neurons project, in the absence of diffusion, GP PV + neurons could nevertheless be retrogradely transduced.
Therefore, we examined the GP of 4 AAV2/9-hsyn and 3 AAV2/1hsyn-injected mice apparently devoided of direct GP transduction as shown in Fig. 2E-H. In one AAV2/9-injected mice, few GFP + cells were observed in the GP (Fig. 6D). As expected these native GFP fluorescent cells were also PV + (Fig. 6E & F). However, the level of fluorescence intensity of GFP + cells in the GP was lower than the fluorescence of directly transduced striatal cells (Fig. 6A), also labeled by anti-PV immunofluorescence ( Fig. 6B & C). Whether these GFP + cells resulted from retrotransduction of pallido-striatal projection neurons or residual diffusion of viral particles remains to be determined. The mean fluorescence intensity of the GFP + cells in the GP (n = 4) was drastically lower than striatal GFP + cells in the same animal (n = 4), respectively 5.90 ± 3.69 and 76.98 ± 7.78 A.U.
In 2 other AAV2/9-injected mice, no GFP + cells were detected in the GP but some fibers were evidenced (Fig. 6G-I). The origin of these GFP + /PV + fibers in the GP remains to be determined. Finally, in the last AAV2/9-injected mice, no GFP + cells of fibers were observed. These data suggest that if AAV2/9 retrograde transduction of pallido-striatal neurons occurred, it was very inefficient.
In AAV2/1-injected mice, devoided of direct GP transduction, no GFP + cells were detected in the GP (data not shown).

Decrease of PV-expressing cells in the striatum of PV Cre mice
Continuous postnatal Cre expression has previously been reported to cause a decrease of cell numbers of some cell types such as immune cells (Schmidt-Supprian and Rajewsky, 2007;Zeitrag et al., 2020) or retinal pigmented epithelium cells (He et al., 2014).
Therefore, we compared the number of PV + cells in WT and PV Cre homozygous mice.
The number of PV + cells was drastically (approx. 2-fold) reduced in the striatum (Fig. 7G) but not in the CTX (Fig. 7H) of PV Cre mice as compared to WT mice (compare Fig. 7A-F).
In order to determine whether Cre effect could be aggravated by viral transduction, injected and non-injected hemispheres of PV Cre mice were compared for the 3 vectors (Fig. 7I). No difference was observed. As a control, the number of PV + cells in the left and right hemispheres of 3 non-injected WT mice were compared. No difference was observed Cre decreases the number of PV + cells in the striatum but not in the cortex. In the striatum (B & E), the number of PV + cells was significantly reduced in PV Cre mice (twoway ANOVA with Sidak's multiple comparison test; p = 0.0201 for AAV2/9-CBA-FLEX-eGFP; p = 0.0064 for AAV2/9-hsyn-FLEX-eGFP; p = 0.0004 for AAV2/1-hsyn-FLEX-eGFP) (G). Circles, AAV2/9-CBA-FLEX-eGFP; squares, AAV2/9-hsyn-FLEX-eGFP; triangles, AAV2/1-hsyn-FLEX-eGFP. In the cortex (C & F), the number of PV + cells in PV Cre mice was not significantly different from the number of PV + cells in WT mice (Student t-test) (H). Circles, wild-type mice; squares, PV Cre mice. Viral toxicity The number of PV + cells in the injected hemisphere of PV Cre mice was compared to the contralateral non-injected hemisphere (I). Statistical analysis did not reveal a significant difference (two-way ANOVA with Sidak's multiple comparison test). Circles, AAV2/9-CBA-FLEX-eGFP; squares, AAV2/9-hsyn-FLEX-eGFP; triangles, AAV2/1-hsyn-FLEX-eGFP.

Discussion
In order to study the role of different types of neurons in complex neuronal circuits, it is crucial to have available molecular tools allowing to selectively deliver genetic information in targeted neuronal subpopulations. Furthermore, it is key to avoid transducing neurons with similar functions in other brain regions which could perturb the behavioral and functional outcomes. Thus, both cell type-specific and region-specific targeting are required.
In the present study, we focused on the targeting of transgene expression into mice PV + interneurons of the dorsal striatum. These interneurons, also called "fast-spiking", although representing a very small proportion of striatal neurons (approx. 0.5 %), are key to the coordination of striatal projections neurons which modulate the sensorymotor loop.
We show that, due to the diffusion of viral particles in the brain parenchyma, it is difficult to target the dorsal striatum while avoiding diffusion to the GP. In contrast to the sparse distribution of PV + striatal interneurons, the GP contains a high density of PV + neurons (40 % of the total) which project to the STN, the SNr and the striatum (Saunders et al., 2016). Transduction of PV + GP projection neurons severely compromised the targeting of PV + striatal interneurons. Precise stereotaxic coordinates adjustment in the dorsolateral striatum, AAV1 capsids (diffusing less than AAV9), and the hsyn promoter (having a moderate transcriptional activity) were necessary to avoid direct transduction of GP PV + neurons. Such an extrastriatal PV + neuronal transduction after striatal injections of AAV-FLEX vectors, has been previously described and authors a posteriori excluded from their behavioral analysis the animals showing a transduction in untargeted areas .
None of the vectors provided any GFP + cell in WT mice demonstrating the tight Cre-dependency of transgene expression. With all tested vectors, the proportion of GFP + cells stained by anti-PV antibodies was higher than 85 %. However, GFP + cells not expressing PV at a detectable level were observed. This could be due to an effect of the Cre recombinase towards PV + neurons. Indeed, we observed that the number of PV + cells was drastically reduced (~ 2-fold) in the striatum of PV Cre mice. This Cre-mediated effect was specific to striatal PV + neurons since no reduction of PV + cells was observed in the cerebral cortex. A similar phenomenon has been reported in retinal pigmented cells in which Cre expression resulted in age-and dosage-dependent attenuation of β catenin and phalloidin stainings in VDM2-Cre mice (He et al., 2014). It remains to be determined whereas Cre effect on PV + cells is also present in heterozygous Pvalbm1/Arbr mice. During the writing of this manuscript, Enteria-Moralez and collaborators reported a similar effect of Cre towards PV + cells in the striatum of homozygous Pvalb-m1/Arbr mice. In this study, heterozygous Cre (+/-) mice showed a non-significant decrease of PV mRNA. However, the small number of mice (n = 2 for some groups) precluded valid statistical analysis.
AAV vectors have previously been reported to have a toxic effect on specific classes of neurons in particular when expressing GFP. Indeed, nigral injection of AAV2/1 vectors in rats resulted in a decrease of the number of TH-expressing neurons, which was further aggravated in the presence of GFP (Albert et al., 2019). AAV2/9 vectors expressing GFP were furthermore shown to elicit an inflammatory response in the rat striatum (Samaranch et al., 2014).
Another potential off-target effect could result from retrograde transport of AAV viral particles in pallidostriatal fibers and transgene expression in GP PV + projection neurons (Saunders et al., 2016). AAV retrograde transport depends on the capsid serotype and inconsistent data have been reported. For example, AAV1 retrograde transport has been repeatedly found to be negligible (Burger et al., 2004;Bockstael et al., 2008;Oh et al., 2014;Tervo et al., 2016) although another report shows that it occurred with a moderate efficiency in nigrostriatal dopaminergic neurons (Taymans et al., 2007). Surprisingly, efficient retrograde infection was demonstrated with AAV2/1-FLEX vectors in several types of neurons (Rothermel et al., 2013) whereas, in the same study, a constitutive AAV2/1 vector expressing mCherry from the CBA promoter but not one expressing eGFP from the hsyn promoter, was proficient for retrograde transduction. Similarly, inconsistent reports of AAV9 serotype retrograde transport efficiency have been published (Klein et al., 2008;Tervo et al., 2016;Cearley and Wolfe, 2006;Gallagher et al., 2008). In a comparative study, (Tervo et al., 2016) AAV1 ability to undergo retrograde transport was found to be lower than that of AAV9. Different species (rat versus mice), different neuronal pathways, expression levels driven by the chosen promoter and titers of the vector may explain these discrepancies.
To our knowledge, retrograde transduction of GP PV + neurons following intrastriatal AAV injection has never been described. In our study, in the absence of apparent diffusion, only a very low number of GFP + cells were detected in 1 out of 4 AAV2/9-injected mice examined, and their native green fluorescence was lower than at the target site in the striatum. Furthermore, it cannot be excluded that the observed GFP + cells in the GP resulted from a low level of diffusion rather than retrograde transport. In contrast, with AAV2/1 vector, no retrograde transduction of pallidostriatal neurons was observed in our conditions.
In conclusion, caution should be taken when interpreting functional and behavioral data obtained in the PV Cre mice/AAV-FLEX system. First, it is key to avoid transducing PV neurons in other brain regions which will perturb the behavioral and functional outcomes. GP PV + neurons project to the STN, SNr and striatum (Saunders et al., 2016), nuclei which are all involved in the motor loop. Thus, transduction of these neurons might drastically modify the functional effects of transgene expression. Secondly, the herein described Cre-mediated deleterious effect toward PV + striatal interneurons could reflect neuronal cell death or loss of PV expression. In the latter case, these neurons are probably dysfunctional since PV, a calcium-binding protein is important for neuronal activity. Perturbation of the synaptic plasticity of these fast-spiking neurons which are thought to coordinate the activity of striatal projection neurons, might severely affect the motor loop. It should be noted that, in a model of AAV-mediated Cre expression, neuronal programmed cell death inducing behavioral perturbations were described (Rezai et al., 2019) In order to achieve targeted gene delivery into fast-spiking striatal interneurons without the use of PV Cre mice and AAV-Flex vectors, a more selective transcriptional targeting could be designed. Viral vectors with specific enhancers allowing to target neuronal or glial cell populations in wild-type animals are coming to an era (Dimidschstein et al., 2016;Meunier et al., 2016;Pignataro et al., 2017;Vormstein-Schneider et al., 2020).

Declaration of Competing Interest
The authors report no declarations of interest.