Serotype-dependent transduction efficiencies of recombinant adeno-associated viral vectors in monkey neocortex.

Abstract. Viral vector-mediated expression of genes (e.g., coding for opsins and designer receptors) has grown increasingly popular. Cell-type specific expression is achieved by altering viral vector tropism through crosspackaging or by cell-specific promoters driving gene expression. Detailed information about transduction properties of most recombinant adeno-associated viral vector (rAAV) serotypes in macaque cortex is gradually becoming available. Here, we compare transduction efficiencies and expression patterns of reporter genes in two macaque neocortical areas employing different rAAV serotypes and promoters. A short version of the calmodulin-kinase-II (CaMKIIα0.4) promoter resulted in reporter gene expression in cortical neurons for all tested rAAVs, albeit with different efficiencies for spread: rAAV2/5>>rAAV2/7>rAAV2/8>rAAV2/9>>rAAV2/1 and proportion of transduced cells: rAAV2/1>rAAV2/5>rAAV2/7=rAAV2/9>rAAV2/8. In contrast to rodent studies, the cytomegalovirus (CMV) promoter appeared least efficient in macaque cortex. The human synapsin-1 promoter preceded by the CMV enhancer (enhSyn1) produced homogeneous reporter gene expression across all layers, while two variants of the CaMKIIα promoter resulted in different laminar transduction patterns and cell specificities. Finally, differences in expression patterns were observed when the same viral vector was injected in two neocortical areas. Our results corroborate previous findings that reporter-gene expression patterns and efficiency of rAAV transduction depend on serotype, promoter, cortical layer, and area.


Serotype-dependent transduction efficiencies of recombinant adeno-associated viral vectors in monkey neocortex 1 Introduction
For more than three decades, viral vectors have been used extensively to deliver genetic material into specific subsets of cells throughout the body, with the ultimate goal of human gene therapy. [1][2][3][4][5] To this end, recombinant viral vectors containing genes-of-interest and free of replication-essential elements have been bioengineered. Presently, such viral vectors are also widely used in basic biomedical research to express proteins in welldefined cell types. 6 In neuroscience, several methods have recently been introduced to control neural activity with unprecedented specificity. Methods such as optogenetics, [7][8][9][10][11] "designer receptors exclusively activated by designer drugs" (in short, DREADD) [12][13][14] and the double-infection technique (DIT), 15,16 rely mainly on viral vectors to introduce genes encoding light-sensitive proteins (opsins), artificial receptors (DREADD), or neurotoxins (DIT) into neurons. Transduction of specific subsets of cells (i.e., tropism) is inherent to the viral vector used and depends on the interaction between viral vector envelope/capsid and receptors on the cell membrane. Subsequent reporter or transgene expression depends on the potency of the heterologous promoter and the availability of transcription factors (TFs) for this promoter in the targeted cell (cell specificity). 17 Despite the abundant use of viral vectors, state-of-the-art neuroscience and clinical studies have exposed potential challenges. These include cross-species variation in vector tissue tropism and gene transfer efficiency, pre-existing humoral immunity to capsids, and vector dose-dependent toxicity in animal models and patients. Detailed information about the transduction characteristics of such vector systems in macaque neocortex, however, is only sparsely available. 18 This is particularly an issue considering the increasing popularity of viral vectors in nonhuman primate research.
The most frequently used viral expression systems are developed from adenoviruses, adeno-associated virus (AAV), herpes simplex virus, gamma retroviruses and lentiviruses (LV; subclass of retroviruses), each having their particular advantages and limitations. 1,19 AAVand LVare most often used in the central nervous system (CNS) since they are relatively easy to produce and evoke minimal immune responses. 17,19 The packing capacity of AAV is low compared to LV, and genes delivered using AAV vectors do remain episomal in the transduced cell. Although LVs showed long-lasting expression 2,20 and were clinically successful in several primate studies, 21,22 AAV is considered to be more appealing for transgene expression in CNS. AAV is small (20 to 25 nm), nonenveloped, single-stranded DNA parvoviruses. 23 They are known for their efficient neuronal (NeuN) transduction, long-lasting transgene expression (up to 2 years in rodents and 6 years in primates) and safety profile (no integration in host-cell genome). 3,[24][25][26][27][28] In addition, they can transduce both dividing and nondividing cells. The best-characterized AAV serotype, first cloned into bacterial plasmids, is AAV serotype 2 carrying an AAV2-derived genome (AAV2/2). 29 Through cross-packaging, AAV2-based genomes can be packaged in alternative AAV serotypes (e.g., rAAV2/5), which yields higher transduction efficiencies than AAV2/2. AAV-based vectors are currently the most widely used for preclinical research, due to favorable characteristics such as higher transduction efficiencies and greater dispersion. [30][31][32] Unfortunately, the latter properties are species specific and even brain region specific. 30,33 Moreover, information about transduction and expression properties obtained in rodents does not always translate to primates. [34][35][36][37][38] In macaque monkeys, a comparison between the transduction properties of different rAAVs in subcortical structures, including the substantia nigra, striatum, 39,40 and basal ganglia, 41 revealed that rAAV2/1 and rAAV2/5 were most efficient when under the control of a synthetic chicken ß-actin promoter. In addition, another study of Sanchez et al. 40 revealed a three-to fourfold higher efficiency of rAAV2/8 compared to rAAV2/1, 2/2, and 2/5 when using a cytomegalovirus (CMV) promoter. Optogenetic studies in macaques confirmed that viral vector transduction in cortex is heterogenous across cortical layers. 35,42,43 Recently, an exhaustive comparison of transduction efficiencies of various recombinant adeno-associated viral vector (rAAV) serotypes and promoters in marmoset, mouse, and macaque neocortex again revealed distinct transduction patterns depending on the serotype and promoter used in the viral vector. 18 Serotype rAAV2 showed a smaller spread compared to rAAV1, 5, 8, and 9, while constructs with the SynI promoter showed higher transgene expression in layers 3 and 5 compared to CaMKII-constructs.
Therefore, to optimize future optogenetic, DREADD, and DIT studies in macaque monkeys and to extend the results obtained by Watakabe et al., 18 we determined the transduction patterns of several rAAV serotypes using the CMV, two variations of the CaMKII, and the enhSyn1 promoters in macaque cortex. Specifically, we investigated the efficiencies, laminar expression patterns, spread, cell specificities and potential antero-and/or retrograde projection properties of 12 different rAAV vector/promoter combinations injected into primary visual and prefrontal cortex of rhesus monkeys, using immunohistochemical evaluation of the reporter genes.

Subjects
Four hemispheres from the forebrains of two adult male rhesus monkeys (Macaca mulatta) were used in this study (M10: 6.2 kg and M37: 6.5 kg). Prior to these terminal experiments, both monkeys had participated in a series of electrophysiological and/or contrast-agent enhanced functional magnetic resonance imaging experiments unrelated to the present study. 44 All experimental procedures were approved by the regulations of the biosafety and ethical committees at KU Leuven and strictly adhered to the guidelines of the National Institutes of Health and the European guidelines for the care and use of laboratory animals.
All vector constructs were generated by the Viral Vector Core at KU Leuven according to published procedures. 32 Briefly, serum-free rAAV production was performed by triple transient transfection of HEK 293T cells (ATCC, Manassas, Virginia) using the AAV transfer plasmids, AAV rep/cap (2/1, 2/5, 2/7, 2/8, or 2/9) and pAd.DELTA.F6 plasmid in the ratio of 1:1:1 using polyethylenimine. The supernatant was harvested 5 days after transient transfection, concentrated using tangential flow filtration, and purified using an iodixanol step gradient. Gradient fractions were collected in 250 μl aliquots and fractions with refraction indices between 1.39 and 1.42 were pooled. The iodixanol of the pooled fractions was exchanged five times with phosphate-buffered saline (PBS), in a Vivaspin 6 (PES, 100,000 MWCO, Sartorius AG, Goettingen, Germany). The resulting concentrated rAAV was aliquoted and stored at −80°C.
The functional titers of rAAV2/7-CaMKIIα 0.4 and rAAV2/ 7-CaMKIIα 1.3 were determined as described earlier. 46 Briefly, 75.000 HEK-293T cells were incubated with a serial dilution of the respective rAAV vectors for 72 h. Following transduction, cells were washed, trypsinized, and the cell pellet was washed three more times with PBS. About 100 ng of genomic DNA was used for the determination of the rAAV genomic copies using a real-time PCR, as described earlier. 33 Each dilution of the viral vectors was used three times and this experiment was done in two different assays.
A univariate general linear model test was performed with SPSS v16.0 for Windows to compare the effects of the dilution and promoter sequence (CaMKIIα 0.4 or CaMKIIα 1.3) on the functional titers, with the assay number as a cofactor. Bonferroni posthoc tests were performed to determine specific differences between the groups. The significance level was set to p < 0.05 for all statistical analyses.

Vector Injections
For the cortical injections of the vector, monkeys were first sedated with Domitor and Ketalar (50 mg∕kg, intramuscular, Pfizer) and anesthetized with isoflurane (AE1.2%). Postoperatively, the animals were treated with analgesics (Contramal, i.m.) and antibiotics (Kefzol, i.m). Five separate series of injections were performed, two in M10 (separated by 7 weeks) and three in M37 (separated by 6 and 3 weeks, respectively; Table 1). The time between the last series of injections and the perfusion was 4 weeks for both subjects. Four series of injections targeted area V1 and one series prefrontal cortex (area 46). Craniotomies were performed above area V1 (M10 and M37) and above frontal cortex (only in M37). After incising and partially retracting the dura mater, viral vectors were injected using a Nanoject II Auto-Nanoliter Injector (Drummond Scientific Company) attached to a custom-built micromanipulator with three translational degrees of freedom (XYZ-manipulator). Next, the vector was loaded into pulled glass micropipettes (outer diameter of 50 μm). At each location, a total volume of 920 nL was injected at 10 equally distributed positions between the gray-white matter boundary and the pial surface (total distance of 1.8 mm) with a series of four injections of 23 nL each at a single position every 10 s and applying an interval of 2 min between consecutive vertical positions [ Fig. 1(a)]. To match the vector to its corresponding position, a photo was taken and distances to main anatomical landmarks (e.g., sulci) were measured at each injection site [see Fig. 1(b) for schematic overview of the injection sites relative to sulci and the border of the respective craniotomies]. Once the viral vector injections were completed at a particular site, the dura mater was sutured, the original bone was used to cover the craniotomy, and the skin was closed over the bone flap.

Perfusion
Four weeks after the final injection series, both monkeys were deeply anesthetized (ketalar, 50 mg∕kg, intramuscular, Pfizer) and euthanized using an overdose of pentobarbital (intravenous, 30 mg∕kg, Sigma). The animal was then transcardially perfused using isotonic saline [9% NaCl in 0.1 M phosphate buffer, pH 7.4 (PB)], followed by 3.5% paraformaldehyde in PB solution and finally with 5% glycerol in PB for cryoprotection. Subsequently, the brain was removed from the skull, cut in six to eight pieces and stored for 3 days in 10% and 3 days in 20% glycerol-PB solution. Tissue pieces were quickly frozen in 2-methylbutane (Merck) at −30 to −40°C, stored at −80°C, and sectioned with a cryostat at −30°C (50-μm-thick coronal sections). The sections were stored as free-floating sections in 0.01% sodium azide containing PBS buffer (0.15 M, pH 7.4) at 4°C until further processing.

Histology
Since the pipette tracks caused by the injections were not visible in unstained sections, a series of cryosections (every 200 μm) were mounted onto glass plates. Cresyl violet and immunohistochemical fluorescent protein (FP) staining were then performed on two different series. Based upon the FP-staining, the injection sites could be determined as the sections with the highest proportion of FP-labeled cells across all layers. Parvalbumin (PV) and NeuN stainings were additionally performed on a series of adjacent sections to calculate the percentage of FP+ cells compared to the total number of neurons (NeuN) and to ascertain cell-type specificity (PV).

Cresyl violet staining
Free-floating sections (50 μm) were mounted onto poly-Llysine-coated (Sigma-Aldrich) glass slides and kept at 37°C for 1 day (Heraeus, type B5050). Cresyl violet staining was then performed according to standard methods to delineate areal borders and cortical layers delineation, as previously described. 47 Table 2 summarizes all details concerning antibodies, dilutions, and buffers used in immunohistochemistry. All free-floating sections (50 μm) were washed twice in 1× tris-buffered saline  (Table 2) sections was rinsed in wash buffer. The tissue was then incubated at room temperature (RT) for 2 h with appropriate secondary antibodies ( Table 2). For biotinylated secondary antibodies, slices were additionally washed (3 × 10 min) and incubated with streptavidin-Cy5 at RT for 2 h (1:500, invitrogen). Following 3 × 10 min washing steps in wash buffer, the sections were rinsed quickly in aqua destillata and sections were incubated for 30 min in 4′,6-diamidino-2-phenylindole (2 μl∕100 ml 0.1 M PBS). Finally, the floating slices were mounted onto gelatin-bloom coated slides (0.05% gelatin bloom solution, type 2, sigma) and cover-slipped with mowiol mounting medium for examination by fluorescence microscopy.

Image Acquisition
All (immuno-)histochemically stained sections were imaged (10×, 20×, and 40× dry lenses) with a Zeiss Axio-Imager microscope equipped with Zeiss CCD Axiocam black and white and color cameras (MRc3 and MRc5, respectively). Fluorescent photographs for semiquantification were additionally acquired with an Olympus Fluoview FV1000 inverted confocal microscope (4×, 10×, and 20× objectives) and using the Olympus Fluoview acquisition software 3.0a. After defining the region of interest, the software divided this region into four to six sections and a z-stack (to cover the 50-μm-thick slice) was acquired.

Proportion of Transduced Neurons
To determine the fraction of all neurons and inhibitory-specific neurons transduced for each vector, confocal images of the NeuN stainings and PV stainings were taken. To approximately define the proportion of colocalizing cells, a semiquantitative analysis (i.e., manual quantification in one section) was performed. Therefore, images were first cropped to the preferred size. Layers were then defined based on NeuN stainings and compared to neighboring CV-stained slices. Afterward, the overlap with fluorescently labeled neurons was manually scored and quantified across the different cortical layers for each image, as described in Diester et al., 35 using ImageJ (image processing and analysis in Java, National Institute of Health).
When a percentage of transduced neurons is mentioned in the results, these numbers were obtained from sections adjacent to the injection site. The percentage was calculated as the number of colocalized FP+ and NeuN+ cells divided by the total number of NeuN+ cells multiplied by 100. The numbers representing the spread are based on the maximal distance from the injection site at which FP+ cells were found across the cortical surface. The anterior-posterior (A-P) axis is always mentioned first followed by the spread along the lateral-medial (L-M) axis.

Results
We injected a series of rAAV vectors into two cortical brain regions of both hemispheres in two monkeys (M10 and M37) to determine their transduction efficiency and apparent distribution pattern, as evaluated by reporter gene expression ( Fig. 1). In M10, serotypes rAAV2/5, 2/7, 2/8 and 2/9 were injected into left and right area V1. FP-expression was driven by i) a constitutive CMV promoter (rAAV2/5, 2/7, 2/8 and 2/9), ii) a long, or iii) short version of the excitatory cell typespecific CaMKIIα promoter (rAAV2/7). The long version of the CaMKIIα promoter (1.3 kb; CaMKIIα1.3) has been used previously and was found to transduce excitatory cells in several species including mouse, rat and monkey. 43,[49][50][51] Additionally, a shorter CaMKIIα promoter (0.4 kb; CaMKIIα0.4) 52 was constructed to facilitate packaging of larger genes (e.g., opsins) in future experiments. In the second monkey (M37), three injection sessions were carried out. In the first session, the rAAV2/1, 2/8 and 2/9 CMV  vectors and a rAAV2/7-CaMKIIα0.4 vector were injected into dorsolateral prefrontal cortex to confirm results obtained in the first monkey. In a second session, performed 6 weeks later, different serotypes (rAAV2/1, 2/5, 2/7, 2/8 and 2/9) under control of the same promoter (CaMKIIα0.4) were injected into left area V1 to compare transduction efficiencies of these serotypes with the results obtained after injections in rodent V1. In the final series of injections (3 weeks later), vectors with the same serotype (rAAV2/7), but with different promoters (CMV, enhSyn1, CaMKIIα0.4 and CaMKIIα1.3), were injected to reveal the effect of the promoter on the transgene expression. Identical injections in V1 and prefrontal cortex (area 46) were designed to compare transduction patterns across two entirely different neocortical areas. Following a description of the FP expression at the injection sites, we will give an overview of retro-and anterograde expression patterns in the dorsal lateral geniculate nucleus (dLGN). Finally we will provide a semi-quantitative overview of the cortical FP expression patterns in a final summary figure.

Recombinant Adeno-Associated Viral Vector
Serotype-Dependent Laminar Transduction Patterns EGFP expression was also less pronounced in layers IVa and IVb (Fig. 3).
Injection of V1 with the rAAV2/7-CaMKIIα0.4 construct yielded FP+ cells mainly in the superficial and deeper layers in both monkeys (Fig. 4). The largest fraction of transduced cells was observed in layers II/III (Fig. 4, panels c and h), followed by layer V and layer VI (Fig. 4, panels e and j). Although neuropil was FP+ in layers IVb, IVc, and IVc, these layers contained relatively few FP+ cells. Hence, this vector was less than optimal for NeuN transduction in the granular layers.
Injection of rAAV2/8-CaMKIIα0.4 resulted in FP+ cells through all cortical layers, with the highest fraction in the deeper layers (Fig. 5). The spread of FP-labeled cells was rather limited along the A-P and L-M axes with a slightly wider spread in the deepest layers (1.8 × 2.0 mm). In addition, many FP-positive glial cells were detected, most notably in the granular layers, making this serotype less specific for NeuN transduction [ Fig. 5(c)].
Injection of serotype rAAV2/9 with a reporter under control of the CaMKIIα0.4 promoter transduced relatively few cells in the supragranular and granular layers in contrast to infragranular layer VI (Fig. 6). The fraction of FP+ neurons was similar to the total number of transduced cells making it better for NeuN expression in comparison to rAAV2/8 but less optimal than rAAV2/7. The spreading of FP-expression induced by this vector is again rather limited (layer VI: 1.2 × 2.0 mm 2 ). Thus, when the CaMKIIα0.4 promoter is injected into monkey striate cortex, the number of cells that express the transgene in V1, the spread from the injection site, and the expression in the various cortical layers depend on the serotype.

Promoter-Dependent Transgene Expression after rAAV2/7 Transduction
We next compared transgene expression driven by different promoters while using the same rAAV serotype. We chose the rAAV2/7 serotype because the previous test yielded a high ratio of transduced neurons in both supra-and infragranular layers and also because results of a previous mouse study using exactly the same constructs as the present study showed high transduction efficiency in area V1. 23 We injected a series of rAAV2/7 vector constructs containing the CMV, enhSyn1, CaMKIIα0.4, or CaMKIIα1.3 promoters. These rAAV2/7 constructs were injected into V1 of both animals, with the exception of enhSyn1 (M37 only).  The transduction pattern induced by the rAAV2/7-CaMKIIα0.4 vector has already been described above (Fig. 4). Surprisingly, when the longer version of the CaMKIIα promoter was used (rAAV2/7-CaMKIIα1.3), a remarkably different laminar expression pattern emerged (Fig. 7). In addition to transducing cells in layers II/III, V, and VI, as with the rAAV2/7-CaMKIIα0.4 vector, the rAAV2/7-CaMKIIα1.3 vector also strongly transduced cells in layers IVa and IVb [Figs. 7(d) and 7(i)]. Furthermore, in contrast to the short CaMKIIα0.4 promoter, hardly any FP+ neurons were found in layer IVcβ (compare Figs. 4 and 7). Overall, the long CaMKIIα promoter resulted in a more efficient transgene expression and labeled cells at greater distances from the injection site than the short promoter (A-P × L-M; layer VI: ½7.7 AE 3.2 × ½4.8 AE 2.5 versus ½2.9 AE 1.9 × ½2.8 AE 0.4 mm; mean AE SD of both monkeys for the long versus short CaMKIIα promoter).
The rAAV2/7-CMV vector used in the present study displayed reasonable reporter levels throughout primary visual cortex (Fig. 8). Moreover, EGFP expression was lowest in layers II/ III [ Fig. 8 Fig. 8(e)]. The majority of the EGFP+ cells after rAAV2/7-CMV injection, however, were non-neural [ Fig. 8(b) inset]. In M10, only a few PF+ cells and fibers were detected and no transduced neurons could be found; hence, this monkey was not used for illustrating this vector.

(c)] and maximal in the deeper layers [
In the second monkey, we injected a vector for which the enhancer region of a CMV promoter 53 was fused with the human Synapsin promoter 35,54,55 (i.e., enhSyn1). This promoter in a rAAV2/7 vector yielded a relatively homogeneous transgene expression across all cortical layers (Fig. 9) and resulted in FP+ cells at larger distances from the injection site compared to the rAAV2/7-CMV vector. Furthermore, a semiquantitative analysis revealed that the fraction of FP+, non-neural cells was very low (<2%) compared to the CMV promoter (>90%), which renders the former vector more suited for transduction in V1.

Cell Specificity of a Long and Short Version of the CaMKIIα Promoter
Viral vectors are the most efficient means with which to transfer genetic material into the central nervous system. Specific expression of a reporter or a therapeutic protein is conditional on the combination of serotype and promoter. 17, 56 Here, we tested whether the CaMKIIα promoter is specific for excitatory cells, as is generally assumed. 42,43 Therefore, we performed double stainings between FP+ cells and the inhibitory cell marker (PV), which is expressed by 75% of the inhibitory cells in primate striate cortex. 57 We calculated the number of FP+ neurons that were also PV+ after injection of two rAAV2/7 vectors driven by either a short or a long CaMKIIα promoter. for each layer, we found that up to 29% of the FP+ cells in layer IVcα are PV+, 17% in IVcβ, and 13% in layer V. For the other layers, the degree of overlap was below 2%. Since the highest proportion of PV+ cells is found in layer IVc, 57 this might explain why we found the highest percentage of FP+PV+ neurons in this layer. Hence, the CaMKIIα0.4 promoter is more specific for excitatory cells than the longer CaMKIIα1.3 promoter.

No overlap was found between FP+ and PV+ cells after
Although genomic titers were comparable between the two vectors, differences between CaMKIIα0.4 and CaMKIIα1.3 in their specificities for excitatory cells could be due to different functional titers. For this reason, the functional titers of rAAV2/7-CaMKIIα0.4 and rAAV2/7-CaMKIIα1.3 were determined by dilution assays (Fig. 11) (see also Ref. 23). The dilution of the viral vector solution influenced the measured titers (p < 0.001) but not the genomic titer of the vector construct (p ¼ 0.152). Thus, because the same dilutions were injected for these vectors, differences in functional titers cannot explain the differential specificity for excitatory neurons in these two promoter sequences.

Projections to the Dorsal Lateral Geniculate Nucleus
One generally assumes that rAAV viral vectors do not transfer trans-synaptically. 58 Thus, FP+ cell bodies found at greater distances from the injection site are labeled retrogradely via transduced fibers entering the injection site. The retrograde capacity of rAAV2 vectors was previously shown by Towne et al. 59 and Zhang et al. 60 Since the dLGN is bidirectionally connected with area V1, we hypothesized that this thalamic nucleus  . Panels e and f reveal EGFP+ fibers, while the fibers in panels g and h are mCherry+. Since more than one of the vectors injected in each hemisphere contained the EGFP or mCherry fluorescent tag, we relied on the detailed retinotopic location of the respective injection sites in V1 and transduced fibers in the LGN to define which vector had transduced these fibers. Based on the corresponding retinotopic locations (eccentricity and polar angle) of the injection sites, we inferred that the band of fibers in panel e was transduced by the rAAV2/8-CaMKIIα0.4 vector and those in panel g by the rAAV2/5-CaMKIIα0.4 vector (approximate eccentricity and polar angle values of, respectively, 6 to 8 deg and þ15 deg and 2 to 3 deg and þ30 deg). The latter vector also resulted in a relatively widespread cortical transduction pattern in layer VI of V1, which can explain the broad signal corresponding to fibers projecting to the dLGN. For the right dLGN of M37, we again used retinotopic criteria to determine that the observed EGFP+ fibers are transduced by the rAAV2/7-CaMKIIα0.4 vector [eccentricity and polar angle of 4 to 6 deg and þ15 deg; Fig. 12   Thus, injection of rAAV2/7-CaMKIIα0.4, rAAV2/7-CaMKIIα1.3, rAAV2/7-enhSyn1, rAAV2/5-CaMKIIα0.4, and rAAV2/8-CaMKIIα0.4 in V1 resulted in anterogradely labeled FP+ axon terminals in the LGN, whereas retrogradely labeled FP+ cell bodies were found only after rAAV2/7-CaMKIIα1.3 vector injections.

Comparison of rAAV2/7-Mediated Transduction Patterns in V1 and Dorsolateral Prefrontal Cortex
First, to determine whether the transduction efficiencies and expression patterns described in the present study are specific for V1 or are similar in different cortical areas, one vector (rAAV2/7-CaMKIIα0.4) was injected in area V1 and in the dorsolateral prefrontal cortex (area 46) of one monkey (M37). Transduction was more homogeneous across all layers of area 46 compared to V1 (Fig. 13), but the area of lateromedial spread in layer VI was half of that in area 46 (2.4 mm versus 5.0 mm). The anteroposterior spread, however, was equal in the two areas (3.0 mm).

Comparison of Cytomegalovirus-Mediated Transduction Patterns in V1 and Dorsolateral Prefrontal Cortex
Finally, to confirm that the absence of FP+ neurons after injection of the CMV-promoter-driven constructs is not restricted to V1, the same vectors were also injected into area 46. All injected CMV-constructs (rAAV2/1-CMV, rAAV2/8-CMV and rAAV2/ 9-CMV) failed to show transgene expression in prefrontal cortical neurons in macaque neocortex.

Promoter-Dependent Transgene Expression Levels
The serotype of the vector determines which cells can be transduced by the viral particles. Furthermore, the promoter should also be active in expressing the reporter protein in these cells.
Thus, the combination of serotype and promoter determines the transduction pattern. Despite the widespread use of the CMV promoter in rodents, 32,33,[64][65][66] we were unable to transduce neurons efficiently with rAAV carrying the CMV promoter in primate cortical neurons, in both visual and prefrontal cortices.
Recently, Watakabe et al. 18 reported neural toxicity caused by the CMV promoter when survival time after injections exceeded 21 days (in our case between 29 and 93 days). In line with this study, we also found more than 90% of the transgene expression driven by rAAV2/7 CMV in glial cells instead of neurons.
In contrast to results with the CMV promoter, the human synapsin-1 promoter 11,35,55 combined with the enhancer elements of the CMV promoter (enhSyn1) yielded a homogenous expression pattern across all cortical layers. The hSyn1 promoter sequence efficiently recruits TFs in monkey neurons based on the high expression levels driven by the enhSyn1 promoter in these cells. In rodents, NeuN TFs have been shown to have a high affinity for the hSyn1 promoter. 63,64 Adding the CMV enhancer sequence (enhSyn1) additionally improves the neuron-specific transcriptional activity of the hSyn1 promoter, in line with previous reports. 67,68 Another surprising finding was the different transduction patterns induced by two highly similar viral vector constructs consisting of different lengths (364 bp and 1291 bp) of the CaMKIIα promoter. We found that in V1, the short 0.4 kb CaMKIIα-promoter transduces neurons in all layers, with the least efficiency in layers IVb and IVcα. The long version CaMKIIα-promoter selectively transduced cells in all cortical layers, however, few neurons in layer IVcα and IVcβ. Thus, the main distinction between the transduction efficiencies of these two versions of the CaMKIIα promoters can be found in the granular layers. The laminar expression of CaMKIIα mRNA in primate V1 has been well described [69][70][71] but fails to explain the differences we observed between the transduction patterns of the two versions of the CaMKIIα promoters used in this study. For instance, layer IVb is almost void of CaMKIIα mRNA signal, 69,71 which fits with the low number of transduced cells in layer IVb using the CaMKIIα0.4 promoter but not with the reasonably high transduction levels using the CaMKIIα1.3 promoter. Based on the lack of expression in specific layers using the CaMKIIα1.3 promoter, we speculate that factors that can inhibit transcription, such as specific miRNAs, may bind to the extra fragments in the longer version of the CaMKIIα promoter. Conversely, higher expression levels induced by the CaMKIIα0.4 promoter may result from the loss of this inhibitory regulation. Irrespective of the reasons for transductional disparities between the short and long CaMKIIα promoters, the former is more specific for excitatory cells when used in combination with AAV.
Other than the specific laminar expression patterns of the two CaMKIIα-promoters, differences in cell specificity were also observed. The rAAV2/7-CaMKIIα0.4 vector resulted in transgene expression exclusively in excitatory neurons while, somewhat surprisingly, rAAV2/7-CaMKIIα1.3 also drove expression in a small proportion of V1 PV-positive cells. Despite the relatively small fraction of PV+FP+ neurons across the cortical depth as a whole, in some layers (e.g., IVcα) almost one third of the transduced neurons were inhibitory in nature. The latter results are contrary to findings in most nonhuman primate 42,43 and rodent studies, 49,51,52,72 where selective expression in excitatory cells was reported when the CaMKIIα promoter is used. A study by Nathanson et al. 46 already showed the lack of specificity of the CaMKIIα promoter when used in an AAV vector, and recent results in mice have confirmed this finding. 50 Furthermore, these authors reported that the specificity of lentiviral vectors with the CaMKIIα promoter is largely due to the tropism of the lentivirus, in which the CaMKIIα can provide additional selectivity. Also in line with our results, the fuguderived CaMKIIα promoter in an rAAV2/1 vector drives expression in a small fraction of inhibitory neurons. 73 In that study, the percentage of excitatory positive-labeled cells was much higher, exactly as in the present study.

Retrogradely Transduced Cells
Previous studies showed that rAAVs can transduce cells retrogradely in nonhuman primates 39,58,74,75 and rodents. 30,32,[76][77][78][79] We confirmed this observation by showing EGFP-labeled cell bodies in the dLGN at locations corresponding retinotopically to the V1 injection sites of the rAAV2/7-CaMKIIα1.3 vector. The degree of retrograde labeling was rather low. Injections of the rAAV2/7-CaMKIIα0.4, rAAV2/5-CaMKIIα0.4, rAAV2/ 8-CaMKIIα0.4, and rAAV2/7-enhSyn1 vectors in V1 resulted only in anterogradely FP-labeled axons in retinotopically matching regions of the dLGN. The differences in retro-and anterograde transduction capacities in the dLGN may be related to the laminar variations in transduction observed in area V1 using the various vectors in the present study. Thalamocortical axons terminate mainly in layer IV; 80 hence, good transduction of that layer should be required for optimal retrograde labeling in the LGN. Conversely, feedback from V1 to the LGN mainly originates in layer VI; thus, efficient transduction in the deepest layer after V1 injections must be a prerequisite for good anterograde labeling in the thalamus.

Transduction Efficiencies are Area Dependent
We found laminar differences in the transduction patterns in two neocortical areas (V1 and prefrontal cortex) of the same monkey, even when injecting the same viral vector (Fig. 13). Hence, as expected, the transduction pattern of rAAV vectors also depends on the targeted area. This suggestion is corroborated by the apparent discrepancy between our results and those obtained in primate SN using the same serotype. 30 The lack of FP+ neurons after injecting a CMV construct, however, was observed in the two neocortical areas. This could indicate that the aforementioned neural toxicity caused by CMV-driven constructs affects all cortical regions.

Titer does not explain differences in neuronal transduction patterns
In the present study, a similar volume was injected for all vector constructs. Several of the viral vector constructs, such as rAAV2/7-CaMKII0.4, rAAV2/8-CaMKII0.4 and rAAV2/9-CaMKII0.4, had almost equal genomic titers (2.6, 2.8, and 3.1 E12 gc∕ml, respectively). These constructs, however, resulted in markedly different numbers of transduced cells and yielded pronounced differences in the spreads of the transduction patterns along the cortical sheet (see Fig. 14). The functional titer essay also argues against genomic titer being an explanatory factor for the observed differences in FP expression patterns (Fig. 11). In addition, similar results with identical viral vectors were found in the mouse visual cortex. 23 We, therefore, suggest that the main factors driving the variations in transduction efficiencies are the serotype, the promoter, and the injected cortical area.

Sample size
Although this study provides useful insights into the transduction capacity of viral vectors in primates, it has its limitations. Sample size was low (i.e., multiple injections were only performed in one animal) and differences in time between injection and perfusion could have produced variations in transductional profiles. Optimal timing for transduction and expression of rAAV may vary according to viral vector, injection area, and species. 81 Nevertheless, previous results in monkeys showed a stable expression of AAV2 after 30 days. 39 Although a staining for inflammatory responses was not performed in this study, obvious inflammatory tissue damage was not discernable. Correspondingly, a comparison of blood samples taken from M37 before and after injection did not reveal any immune responses.

Conclusion
The present study compared NeuN tropism of rAAVs in macaque cortex. Our results illustrated that different rAAV constructs (various recombinant AAV vectors in combination with different promoters) yield distinct transgene expression patterns in terms of spread, laminar distribution, cell specificity, and retrograde transduction capacities. The comparison of multiple viral vector systems within the present study and across previously published studies revealed that transduction patterns of viral vector systems depend on (1) the serotype, (2) the promoter sequence, and (3) the targeted area. Given the variability in transduction efficiency obtained using various rAAV constructs, we advise performing histological checks in each viral vectorbased experiment. More provocatively, these results indicate that by selecting the appropriate combination of vector and promoter, one may target specific NeuN circuits within primate striate cortex, confined to specific layers and types of neurons.
Pascaline Vancraeyenest obtained her MA at the biomedical sciences department of KU Leuven. She conducted her master's thesis under the supervision of Wim Vanduffel and Annelies Gerits in the laboratory of neuro-and psychophysiology. Her master's thesis was nominated second-best of her graduating year. She is now a doctoral fellow of the IWT Vlaanderen and is still working in the laboratory of Wim Vanduffel.
Samme Vreysen obtained his master's degree in biochemistry and biotechnology at KU Leuven and is working towards a PhD in the laboratory of neuroplasticity and neuroproteomics of Lutgarde Arckens. Veerle Baekelandt obtained her PhD in biology in 1995 at KU Leuven and then did postdoctoral work at the faculty of medicine. She now has her own research group (Laboratory for Neurobiology and Gene Therapy at the Department of Neurosciences), focusing on familial forms of Parkinson's disease and adult neurogenesis using viral vector technology, stereotactic neurosurgery, and noninvasive molecular imaging as core technologies.

Marie-Eve
Lutgarde Arckens obtained her PhD from the laboratory of neuroplasticity and neuroproteomics at KU Leuven and then worked as a postdoctoral fellow in the same laboratory. Since October 2003, she has been a full professor at KU Leuven and is the head of the research group of neuroplasticity and neuroproteomics focusing on the molecular and structural mechanisms of plasticity in mammalian cortex.
Wim Vanduffel obtained his PhD from the laboratory of neuro-and psychophysiology at the KU Leuven and then worked as a postdoctoral fellow in the same laboratory. Afterwards, he served as instructor to the Martinos Center for Biomedical Imaging of the Massachusetts General Hospital, becoming an assistant professor at Harvard Medical School. He is also currently a full professor at KU Leuven and serves as administrative head of the lab of neuro-and psychophysiolgy.