A new AAV tool for highly preferentially targeting hippocampal CA2

6–8 week-old male C57BL/6 mice were kept group-housed in 4 per cage. Inbred mouse strain Tg (Amigo2-cre) 1Sieg (Amigo2-Cre) was acquired from the Jackson Laboratory and inbred in the Animal Core Facility of Nanjing Medical University (Nanjing, China) for experiments. The mice were housed under standard laboratory conditions with access to food and water ad libitum, stable temperature (22 ± 1 °C), and 12-h light–dark cycle (lights on at 07:00). All animal care and experimental procedures were followed by the Animal Experimental Ethical Guide of Southeast University and Animal Core Facility of Nanjing Medical University.

The promoter of specific CA2 expression gene was identified based on Allen Brain Atlas and Mouse Genome Informatics. The sequences of mitogen-activated protein kinase kinase kinase 15 (Map3k15, NM_001163085.3, NC_000086.7) and regulator of G protein signaling 14 (RGS14, NM_001360714.1, NM_016758.3) were downloaded from the NCBI. Map3k15 and RGS14 promoter were amplified using mouse DNA as a template. Activity of promoters was detected by OIBO company (Shanghai, China).

And the following primers are used for the amplifications: 5′-gctggaactagggaaggatttc-3′ (Map3k15-2.3 kb promoter-F) & 5’-atcgtagaaggcatcgagcac-3’ (Map3k15-2.3kb promoter-R);

5′-ggaggaagtcagggagggaggaaa-3′ (Map3k15-0.5 k promoter-F3) & 5’-gcggggctggcggcttcgaa-3’ (Map3k15-0.5k promoter-R3);

5′-tgctccctatctctctgctttctg-3′ (RGS14-1.4 k promoter-F) & 5’-tctgccacaggaagtgtctcta-3’ (RGS14-1.4k promoter-R)

.

The specific CA2 promoter construction plasmids (rAAV-M1-NLS-CRE-WPRE-pA, M1 can be replaced by the M2 or RGS14 promoter) were packaged into Adeno-associated viruses (AAV)-M1-CRE (Map3k15-0.5 kb), AAV-M2-CRE (Map3k15-2.3 kb), and AAV-RGS14-CRE (RGS14-1.4 kb) by OBIO (Shanghai, China). AAV-CMV-EGFP (AOV022-1 AAV2/9), AAV-hSyn-DIO-mcherry (H4828 AAV2/9 CK1214 AAV2/9) were purchased from OBIO (Shanghai, China). AAV-Ef1a-DIO-EGFP (PT-0795 AAV2/5, PT-0012 AAV2/9), AAV-hSyn-CRE (PT0136, AAV2/9), AAV-Ef1a-DIO-hM4Di-mcherry (PT-0043 AAV2/9) were purchased from VTA (Wuhan, China). Clozapine-N-oxide (CNO, A3317) was purchased from APEXbio (TX, USA).

P30 mice were anesthetized with isoflurane (2–5%) (RWD life Science Co, Shanghai, China) and had their scalp immobilized in a stereotaxic apparatus (RWD). After a precise craniotomy, the AAV was loaded in a glass pipette pulled by a heater, which was slowly targeted for injection into the dorsal hippocampus CA2 (left: A/P: − 1.55 mm, D/V: − 1.70 mm, M/L: − 1.70 mm; right: A/P: − 1.55 mm, D/V: − 1.72 mm, M/L: 1.63 mm). AAV-M1-CRE (2.99 × 10¹² vg/ml) and AAV-hSyn-CRE (1.09 × 10¹² vg/ml) were injected with equal amounts of viral particles by controlling the injection volume, and the volume ratio of AAV-CRE and AAV-DIO was ensured to be 2:3. A syringe pump (KD Scientific, Holliston, USA) was used to inject 100 nl of AAV into the dorsal CA2 region bilaterally at 10 nl/min. The needle was gradually withdrawn 20 min after injection. To prevent hypothermia after anesthesia, the mice were positioned on a heating pad through the procedure until fully awake. The injected mice were monitored daily for recovery status, body weight and infection.

Only 6–8 weeks old male C57BL/6 mice were used in behavioral test. For 3-chamber social test, two stranger mice were housed individually in their cages. The mice placed in the engagement area were age-matched, unfamiliar, and of the same genetic background and gender. To familiarize themselves with being held, the strange mice were acclimated 2 days in advance, 45 min a day, which was to avoid anxiety-induced odor and sounds from unfamiliar mice affecting the subjects’ activities. CNO was intraperitoneally injected at 0.5 mg/kg 20 min before the behavioral test.

The open field apparatus is a blue Plexiglas uncapped box (50 cm wide × 50 cm long × 50 cm high), which consists of two parts, the center (29.7 cm diameter in the center of the box) and the periphery. Mice were allowed to move freely for 10 min in the box. The movement of mice was recorded and analyzed with a webcam and Ethovision XT software (Noduls, Wageningen, Netherlands).

The three-chamber behavior apparatus (60 cm long × 40 cm wide × 22 cm high) containing three compartments with equal size, which were separated by two walls with small doors (5 cm wide × 8 cm high). 20 min after intraperitoneal injection of CNO or vehicles, mice were placed in the device and allowed to explore freely for 10 min. The left and right chambers were placed with empty inverted uncapped cages before the experiment. Stranger mice (stranger 1, S1) were placed in the inverted cages in the chamber where the subject mice were not habitually explored. During the social novelty stage, the time that the test mice spent sniffing the novel mice and the empty cage in the other chamber was compared. After 10 min, the test mice were isolated in the middle chamber for 10 min, and at the same time a second novel stranger mouse (stranger 2, S2) was carefully put into the inverted cage that was originally empty in the previous stage. The subject mice were released from the middle chamber to explore the two compartments, and social preference was determined by comparing the time the mice took to sniff S2 and S1 in the last 10 min. Before introducing new subjects into the apparatus, the chambers and inverted cages were thoroughly cleaned with 25% ethanol. Behavior was monitored and scored for social novelty and memory which were represented by the anogenital and nose-to-nose sniffing time of the subject mice. Sniffing time was calculated by a trained observer, who was unaware of genotype and treatment of the mouse. The formulas, social difference score = Time_stranger1-Time_empty (stage 1), and social difference score = Time_stranger2-Time_stranger1 (stage 2) were used to define the social novelty and social memory of the subject mice.

After appropriate anesthesia with the CO₂ bullets, animals were perfused intracardially with ice-cold PBS followed by 4% ice-cold PFA/PBS. Mouse Brains were fixed in 4% PFA overnight and then transferred into 30% sucrose for dehydration. After being embedded in Tissue-Tek^® O.C.T., the brains were snap-frozen in liquid nitrogen and stored at − 80 °C. The brains were sectioned into the slices with the thickness of 30 µm using a Leica cryostat (Leica CM1950, Wetzlar, Germany).

Brain serial sections were attached to glass slides coated with poly-L-lysine (Cat# P8920, Sigma, Germany), and then washed for 15 min with PBS. Heated-antigen retrieval was performed for staining. In general, water-bathed sections were incubated at 96 °C for 5 min in 50 mM sodium citrate pH 8.5 (P0083, Beyotime, Shanghai, China). After being washed from the incubator for 5 min, the sections were permeabilized in 0.3% Triton X-100 in PBS (PBT) for 30 min. Subsequently, the sections were blocked at room temperature for 2 h with 10% Fetal Bovine serum (FBS) in 0.3% PBT (10% FBST) before overnight incubation with primary antibodies at 4 °C diluted with 10% FBST (1:500). The next day, the sections were washed 7 times for 35 min in 0.3% PBT and stained with secondary antibody (1:500) in 10% FBST for 1 h, followed by 5 times washing in 0.3% PBT for another 25 min. After the sections were stained with DAPI (C1002, Beyotime), the slices were mounted. Images were captured under a Zeiss confocal microscope LSM 700 or LSM 900 (Zeiss, Oberkochen, Germany). The whole coronal brain image was taken by Palm Microbeam (Zeiss).

For RGS14, Calretintin, Calbindin, GAD67, Parvalbumin, NeuN, mcherry, CaMKII-α, PCP4 and GFP staining, the brain sections were incubated with the primary antibodies (mouse antibody against RGS14, NeuroMab, Cat:75-170; rabbit antibody against Calretintin, Proteintech, Cat:66496-1-Ig; mouse antibody against Calbindin Proteintech, Cat:14479-1-AP; mouse antibody against GAD67, Millipore, Cat:MAB5406; mouse antibody against Parvalbumin, sigma, Cat:048M4753V; mouse antibody against NeuN, Millipore, Cat:MAB377; rabbit antibody against mcherry, Invitrogen, AB-213511 rabbit antibody against PCP4, SIGMA, HPA005792; rabbit antibody against CaMKII-α, MERCK, C6974 and rabbit antibody against GFP, Torrey Pines Biolabs, Cat:TP401) and secondary antibodies (Alexa Fluor 488 donkey anti-rabbit IgG, Invitrogen, A21206; Alexa Fluor 488 donkey anti-mouse IgG, Invitrogen, A21202 and Alexa Fluor 555 goat anti-rabbit IgG, Invitrogen, A21429). For CA2 spine imaging, RGS14 and CaMKII-α were costained on 1.5%PFA fixed brain slices with the thickness of 150 μm prepared by a vibratome. Sections were incubated with 5% FBS in 0.1% PBT for 1 h, then incubated with the primary antibodies against RGS14 (1:500) and CaMKII-α (1:500) in 3% FBS (0.05% PBT) overnight, and visualized using secondary antibody Alexa 555 (1:200) and Alexa 647 (1:200). The cells were counted on the complete coronal hippocampus images which were acquired through stitching the 10 × confocal microscopic diagrams together. RGS14⁺ cell number in RGS14 stained region was defined as total CA2 cells (Cell sum) and is normalized as 100%. The formulas for Fig. 2G are shown as below:

ImageJ software (National Institutes of Health, Bethesda, USA) and ZEN (Zeiss, Oberkochen, Germany) were used to count cells and measure soma size.

Mice were perfused with 1.5% PFA, and the brains were postfixed by the same solution overnight. Then, 150 µm-thick sections were prepared by a vibratome and labeled by Dil-labeled Au particles using Gene gun. After 24 h at room temperature proceed with incubation with antibody to RGS14 (1:500) in 10% FBS for 16 h, and visualized using secondary antibody to mouse Alexa 647.

Ordinary one-way ANOVA, post-hoc test Bonferroni’s test, and two-tailed unpaired Student’s test was used to assess statistical significance as described in the figure legends. GraphPad Prism 7 (GraphPad, CA, USA) was used for data analysis. Photoshop CS6 (Adobe, CA, USA) was used to merge confocal images to fully present the hippocampus. All the statistical data in the graphs are presented as the mean ± SEM. p < 0.05 was considered significant. All p values were shown in the figures and n represented the number of animals.

Previous studies reported several CA2 markers such as PCP4, RGS14, Amigo2 and Map3k15 [13, 18, 19]. After scrutinizing the expression pattern of these markers in hippocampus in the literatures and the Allen Brain Atlas database (https://​portal.​brain-map.​org), we decided to choose comparatively more restricted CA2 markers, RGS14 and Map3k15, for further promoter activity and specificity test. We wonder which gene and which promoter region are the best choice for CA2 highly preferential viral plasmid construction. Therefore, several vectors containing different promoter fragments from rgs14 and map3k15 with various lengths were constructed and tested. Using the luciferase assays for the promoter activity tests, we screened out a fragment from map3k15 with the length of around 500 bp, which displays the highest luciferase activity (Fig. 1A). To primarily test if it also has a good specificity, the viral vectors containing different rgs14 and map3k15 promoters fused with a Cre gene (Fig. 1B) were packaged and injected into the dorsal CA2 region with AAV9/hSyn-DIO-mcherry. The AAV9/CMV-EGFP was co-injected with the same dosage as a control. This preliminary study indicated the best CA2 marking effects of AAV9/Map3k15 (0.5 kb)-Cre (M1-Cre) instead of AAV9/Map3k15 (2.3 kb)-Cre (M2-Cre) or AAV9/RGS14-Cre (Fig. 1C–E). Meanwhile, the co-injection results of CA2 highly preferential AAV9/M1-Cre & EF1α-DIO-mcherry system and nonspecific AAV9/CMV-EGFP virus into CA2 showed in Fig. 1C–E suggests that EGFP driven by the CMV promoter is expressed much more pervasive in hippocampus. Although the precise stereotaxic was used, it was still impossible to restrict protein expression in the CA2 region without the CA2 specific or highly preferential promoter. The specificity of M2-Cre in CA2 was slightly more accurate than that of the RGS14 promoter, but they were not as accurate as the M1-Cre.

To further test if our M1-cre can truly target CA2 region, serial sections were collected and thoroughly investigated 3–4 weeks after the stereotaxic coinjections of AAV9/M1-Cre & AAV9/EF1α-DIO-EGFP (or EF1α-DIO-mcherry, AAV5/EF1α-DIO-EGFP) mixture performed as before in dorsal CA2 (Fig. 2A). And RGS14 was used as a CA2 marker to delineate this region. Our results displayed that GFP was efficiently and accurately expressed through dorsal CA2 without leakage by M1 driven Cre expression (Fig. 2B1–B5). Again, the non-CA2-specific promoter hSyn or CMV driven Cre was used as a control. The same dosage of AAV9/hSyn-Cre & EF1α-DIO-EGFP co-injection led to a broader nonspecific expression beyond CA2 region. The expression of non-specific EGFP could be pervasively detected in CA1, CA3 and DG (Fig. 2C1–C5, D1–D5). Contrarily, RGS14 could be much better colocalized with M1-driven EGFP expressed cells (the percentage of EGFP⁺RGS14⁺ cell is ~ 48% of total RGS14⁺ cells) in CA2 (Fig. 2G) but much less overlapped with CMV (the percentage of EGFP⁺RGS14⁺ cell is ~ 13% of the total RGS14⁺ cells) or hSyn driven EGFP expressed cells (the percentage of EGFP⁺RGS14⁺ cell is ~ 3% of the total RGS14⁺ cells) (Fig. 2G). In conclusion, there are much less viral expressing cells outside of CA2 regions when applying M1 instead of nonspecific promoters to drive GFP expression (Fig. 2G). And over 95% of these M1-driven EGFP expressed cells are not colocalized with the interneuron markers (Fig. 2F1–F4, I) which means most M1-driven EGFP expressed cells are pyramidal neurons. Moreover, 84.7% of M1-driven EGFP pyramidal neurons in CA2 region receive the projections from Calbindin positive (CB⁺) mossy fibers from dorsal dentate gyrus (Fig. 2F5, J) which was consistent with the phenomenon observed in PCP4⁺ pyramidal neurons in dorsal CA2 [18]. One thing surprising for our immunofluorescence results is that a considerable number of M1 expressing cells does not colocalize with RGS14 (Fig. 2E1–E4). The ratio of EGFP⁺RGS14⁻ cells within RGS14 stained region (~ 16% of the total RGS14⁺ cells) for M1-Cre treatment (Fig. 2G) to EGFP⁺RGS14⁺ cells within the same region (~ 48% of the total RGS14⁺ cells) (Fig. 2G) is around 1:3, which means around 25% of M1 expressing cells does not express RGS14. This is different with the reported result showing almost 100% coexpression of STEP, PCP4 and RGS14 within CA2 region [18], and the result in another previous study showing ~ 97% overlapping between RGS14 and Amigo2 positive cells in this region [13].

To determine if those M1⁺RGS14⁻ cells are real CA2 cells instead of pyramidal neurons from CA1 or CA3, we first thoroughly re-compare the size of cell bodies in CA1, CA3 and CA2 regions as delineated in Fig. 3A1–A3. There are some of them are in the CA2 Z3-Z4 border or in the CA2 Z2-Z3 area according to previous definition of CA2 (Fig. 3B1–B3) [12]. We found the average size of cell bodies in CA2 regions is obviously larger than CA1 pyramidal neurons (Additional file 1: Fig. S1A). And M1⁺RGS14⁻ neurons from CA2 pyramidal cell (Py) layer also show the larger cell body size than CA1 pyramidal neurons in Py layers as previously reported [20]. M1⁺RGS14⁻ neurons from stratum oriens (Or) layer are also quantified and their body size is close to the body size of PV and CB interneurons but not CR interneurons in Py layer (Additional file 1: Fig. S1A). To further investigate the features of these M1⁺RGS14⁻ cells, another CA2 marker PCP4 was utilized and indicated the high colocalization with RGS14 staining cells (Fig. 3C1–C4, D1–D4). PCP4/M1 and RGS14/M1 have similar non-colocalization ratio in the whole CA2 (Additional file 1: Fig. S1B, C) or in different CA2 layers (Additional file 1: Fig. S1D). Next, we used CaMKII-α to label those M1⁺RGS14⁻ cells in CA2-Z3 or CA2-Z4 regions (Fig. 3B4, E1–E5) and it does demonstrate that they are CaMKII⁺. Sometimes even a large area of CA2 cells can be detected as M1⁺RGS14⁻ CaMKII⁺ cells (Fig. 3F1–F4). Those pieces of evidence highly suggest most of M1⁺RGS14⁻ neurons are pyramidal neurons.

Meanwhile, we carefully observed dendritic spines especially spines in apical dendrites of both M1⁺RGS14⁻ and M1⁺RGS14⁺ neurons in stratum lucidum (SLu) layer of CA2 (Fig. 4A, B1–B2). We found most of M1⁺RGS14⁺ cells have typical CA2 simple spines as shown in Fig. 4B1 and C1, and comparatively much rare complex spines can be detected in this group of cells. However, a different case was found in M1⁺RGS14⁻ cells. Besides pyramidal neurons with simple spines were detected in this group of cells, the neurons with complex spines were also detected especially in CA2-CA3 mixed borders (Fig. 4B2) which used to be more commonly observed in CA3 pyramidal neurons (Fig. 4C2). Since CA2–CA3 has a mixed border, the M1⁺RGS14⁻ neurons with complex spines in this region are possibly the CA3 neurons. Meanwhile they are in RGS14 covered region and stained with CaMKII (Fig. 3F1–F4), we call them CA3-like pyramidal neurons. We quantified the soma size and percentage of neurons containing complex spine and neurons with simple spines in CA2–CA3 border. We found the soma size of M1⁺RGS14⁻ with complex spines in these region is bigger than that of the M1⁺RGS14⁻ neurons with simple spines (Fig. 4G, the bar graph), which does mean M1⁺RGS14⁻ with complex spines in this region are similar with CA3 neurons. And 48.7% of M1⁺RGS14⁻ neurons belong to this situation. Around 51% of M1⁺RGS14⁻ neurons have simple spines and are still normal CA2 pyramidal neurons (Fig. 4G, the pie chart), which are stained with CaMKII (Fig. 3B4). Last, we closely study the morphology of M1⁺RGS14⁻ and M1⁺RGS14⁺ cell bodies. Surprisingly both of them appear diverse shapes from typical pyramidal form (Fig. 4D) to other kind of shapes such as the triangle ones as shown in Fig. 4E1–E3. The complex spines were also observed on the apical dendritic tuft in both of RGS14⁺ and RGS14⁻ spines in SLu layer as shown in Fig. 4F1–F5.

We summarize what we found about CA2 M1⁺ neurons in Fig. 4H, most of M1⁺ cells can colocalized with RGS14 as M1⁺RGS14⁺ neurons (yellow). But some of them are M1⁺RGS14⁻ neurons. They can be detected in different area of CA2. Half of them are normal CA2 pyramidal neurons. But half of them are CA3-like neurons (grey) within CA2–CA3 mixed border (Z4 area) or CA2b region tightly close to the border (Z4–Z3 border). The case to detect CA1-like M1⁺RGS14⁻ neurons in CA1–CA2 border is very rare comparing to the possibility to detect CA3-like M1⁺RGS14⁻ neurons in CA2–CA3 border.

The mixed usage of two viruses usually has a lower site accuracy than the single viral usage. It is consistent with our results in Fig. 5. Compared to a broad diffusion to the whole hippocampus and even the cortex caused by mixed viral usage of AAV9/hSyn-Cre & AAV9/EF1α-DIO-EGFP in the dorsal CA2 (dCA2) region (Fig. 5A–C), the single usage of AAV9/CMV-EGFP has a lower diffusion beyond hippocampus (Fig. 5D–G). But both of these two applications cannot restrict the injected viruses just within dCA2. The application of AAV9/EF1α-DIO-EGFP into CA2 specific Amigo2-Cre mice efficiently help to increase the specificity of EGFP expression in dCA2, while a comparatively low level of leakage into cortex still can be detected (Fig. 5H–K). Surprisingly, our combined administration of AAV9/M1-Cre & AAV9/EF1α-DIO-mcherry achieved the optimal restriction of fluorescent protein in dCA2 region and lowest viral leakage to the peripheral regions (Fig. 5M, Additional file 1: Fig. S2A1-A3, B1-B4, C1-C4) comparing to CMV (Fig. 5L) or even Amingo2 driven EGFP expression (Fig. 5I–K). The small leakage as shown in Additional file 1: Fig. S2 A1 actually used to be mistakenly confused with the neurons in CA2a area in many other reports. The small leakage as shown in the Additional file 1: Fig. S2 A2 was confirmed to mostly take place in the Or layer. And the small leakage in CA3 as shown in Additional file 1: Fig. S2 A3 is very rare.

To test if M1 driven system can be utilized for anterograde neural tracing, AAV9/M1-Cre & AAV9/EF1α-DIO-mcherry or AAV9/M1-Cre & AAV9/EF1α-DIO-EGFP were injected and fluorescent signals were traced (Fig. 6A). We successfully observed projections from CA2 to the following regions: hippocampal commissure and fornix (Fig. 6B, C), contralateral CA1, CA2 and CA3 regions (Fig. 6D, E), ipsilateral CA1 and CA3 (Fig. 6F, G), lateral septum (LS), medial septum (MS), ventromedial hypothalamus (VMH), striohypothalamic nucleus (STHY) and ventral premammillary nucleus (PMV) (Fig. 6I–L). Meanwhile, all of the serial sections for neural tracing were carefully examined to guarantee that there is no obviously discernable leakage as in Figs. 5M and 6M. In our tracing experiments, some known CA2 downstream regions can be constantly recapitulated [16, 21, 22]. But VMH, STHY and PMV are the new projected regions we detected using this M1 system. What we find for CA2 projection tracing further prove that our system can accurately label CA2 and be applicable for CA2 specific neural tracing.

Last we wonder if AAV9/M1-Cre virus suitable to be applied into neural activity regulation through chemogenetic manipulation in CA2 region. To test chemogenetic efficiency, AAV9/M1-Cre & AAV9/EF1α-DIO-hM4Di-mcherry were co-injected into CA2 (Fig. 7A) and successfully expressed after 4 weeks of injection (Fig. 7B–D). Then the neural activity in CA2 was inhibited by CNO induction in the viral injected mice. The mice with CA2 neural inhibition displayed the classical social memory defect in three-chamber social test (Fig. 8A–D) without the defects of social ability, anxiety level or motion ability (Fig. 8E–H).