Caspase-3 and GFAP as early markers for apoptosis and astrogliosis in shRNA-induced hippocampal cytotoxicity

RNA interference (RNAi) is a popular technique for uncovering the function of individual proteins in cells or organisms (Shan, 2010). As short hairpin RNAs (shRNAs) can be designed specifically for any structural gene, they provide an extensive toolbox for RNAi-based manipulation of gene expression. However, some adverse effects of shRNA expression, including reduced cell viability, have been observed (Ehlert et al., 2010; Grimm et al., 2006).

Certain aspects of shRNA-induced cytotoxicity have been described in previous studies (Börner et al., 2013; Bridge et al., 2003; Grimm et al., 2010; Sledz et al., 2003; Yi et al., 2005). Notably, shRNA-induced neurotoxicity is affected by various factors, such as shRNA sequence, shRNA concentration applied, means of application (e.g. virus-mediated delivery and expression), as well as the treated brain region. Whereas shRNA-induced neurotoxicity culminating in tissue atrophy has been described for some CNS regions, including the striatum (Martin et al., 2011; McBride et al., 2008), adverse effects have not yet been described in the mouse hippocampus. Nonetheless, virus-mediated delivery of shRNA-expressing constructs is used increasingly to investigate the contribution of individual proteins to hippocampal function on the cellular and subcellular level. At present, information on molecular markers for shRNA-induced tissue stress, even before apparent tissue degeneration, is scarce. Testing for the expression of such markers would allow assessment of any treatment-induced side effects in behavioral studies using shRNA-based approaches.

Here, we report on shRNA-induced degeneration in the murine hippocampus. We examined tissue sections after rAAV9-mediated expression of shRNA constructs by immunohistochemistry. In samples with high expression of different shRNA constructs, strong degeneration and tissue atrophy were detected. We identified two markers, active caspase-3 (Casp-3) and glial fibrillary acidic protein (GFAP), that indicate adverse side effects of shRNA treatment even before tissue degeneration could be observed.

Sequences encoding the human U6 (hU6) promoter and shRNA constructs directed against two hyperpolarization-activated and cyclic nucleotide-gated channel genes, hcn1 (CCG GCC TCC AAT CAA CTA TCC TCA ACT CGA GTT GAG GAT AGT TGA TTG GAG GTT TTT G, no. TRCN0000103027) and hcn2 (CCG GCC ATG CTG ACA AAG CTC AAA TCT CGA GAT TTG AGC TTT GTC AGC ATG GTT TTT G, no. TRCN0000103001), were purchased from Sigma-Aldrich (Munich, Germany). A construct encoding a Photinus pyralis luciferase (luc)-targeting shRNA sequence (Premsrirut et al., 2011) was designed (CCG GCC GCC TGA AGT CTC TGA TTA ACT CGA GTT AAT CAG AGA CTT CAG GCG GTT TTT G) and synthesized (MWG Operon, Ebersberg, Germany). Constructs were cloned into the rAAV vector pENN-CaMKIIeGFP provided by the University of Pennsylvania Vector Core (see Fig. 1). Viral particles were packaged by the University of Pennsylvania Vector Core and genomic titers ranged from 10⁹ to 10¹⁰ genome copies µl⁻¹.

A total of 10 male C57BL/6J mice (The Jackson Laboratory, Bar Harbor, ME, USA) were used in this study for viral injection and subsequent immunohistochemical analysis. Animals were housed in groups of 4–5 under standard conditions with access to food and water ad libitum on a 12 h:12 h light:dark cycle. Experiments were carried out in accordance with National Institutes of Health guidelines and were approved by the University of Pennsylvania Institutional Animal Care and Use Committee. Mice received bilateral intra-hippocampal injections of rAAV9 vectors encoding enhanced green fluorescent protein (eGFP) and shRNA constructs at 8–10 weeks of age. Stereotaxic injection of 1 µl viral suspension (0.2 µl min⁻¹) per hippocampus was performed at antero-posterior −1.9, medio-lateral ±1.5 and dorso-ventral −1.4 from Bregma using a 33 gauge beveled NanoFil needle, a NanoFil syringe and a MicroSyringe Pump Controller (World Precision Instruments, Sarasota, FL, USA). After surgery, mice were single-housed and given 5 days to recover before pair-housing. Injection and tissue collection were performed during the light phase.

Hippocampal tissue from 1–3-day-old mice (C57BL/6 strain from an in-house animal breeding facility) was prepared in ice-cold HBSS (Hanks’ balanced salt solution). Hippocampi were incubated in papain solution [DMEM (Invitrogen), 25 U ml⁻¹ papain, 1.6 mmol l⁻¹ l-cysteine, 1 mmol l⁻¹ CaCl₂, 0.5 mmol l⁻¹ EDTA] at 37°C for 20 min and subsequently in inactivating solution [2.5% (w/v) trypsin inhibitor, 2.5% (w/v) albumin, in FCS solution] at 37°C for 5 min. Cells were triturated in FCS solution [DMEM, 100 U ml⁻¹:100 μg ml⁻¹ streptomycin:penicillin, 10% FCS, 0.1% (v/v) MITO+ serum extender]. Primary hippocampal neurons (PHNs) were plated on μ-Slide 4 Well slides (Ibidi, Martinsried, Germany) pre-coated with poly-d-lysine (0.2 mg ml⁻¹ poly-d-lysine, 50 mmol l⁻¹ H₃BO₃, 25 mmol l⁻¹ Na₂B₄O₇, pH 8.5). PHNs were maintained in NBA medium [Neurobasal A Medium (Invitrogen), 100 U ml⁻¹:100 μg ml⁻¹ streptomycin:penicillin, 2% (v/v) B27-supplement (Invitrogen) and 1% (v/v) Glutamax] at 37°C, 5% CO₂ and 95% relative humidity for 14 days.

Total RNA was isolated from PHN cultures using the DNA/RNA/Protein AllPrep^® Kit (Qiagen, Hilden, Germany) according to the supplier's protocol. RNA samples were split for two independent first-strand cDNA syntheses using Oligo-dT primers (Qiagen) and Moloney Murine Leukemia Virus reverse transcriptase (M-MLV-RT, Life Technologies) according to the supplier's protocol. Thermocycling was performed in a LightCycler 1.5 (Roche, Mannheim, Germany) using the QuantiTect SYBR Green PCR Kit (Qiagen) according to the supplier's protocol. Gene-specific primers (Table S1) were designed in silico and synthesized by MWG Operon (Ebersberg, Germany). Specificity and efficiency of primers was confirmed via BLAST analysis and semi-quantitative PCR on cloned gene fragments (Figs S1, S2). The gapdh primers were designed to bind in exons separated by an intron of 134 bp to check for genomic impurities.

qPCR reactions were performed on 1 μl aliquots of first-strand cDNA samples in a total volume of 20 μl. Results were analyzed using the Ct method. Gene expression levels were normalized to expression of the housekeeping gene gapdh. For analysis, mean±s.e.m. values were calculated. The two-tailed independent Student's t-test was applied for calculation of P-values. A P-value of <0.05 was considered significant.

For immunohistochemical analysis, brains were dissected after transcardial perfusion with ice-cold PBS followed by 4% paraformaldehyde in PBS under deep anesthesia. Tissues were cryo-protected in 30% sucrose (in PBS) for 2 days, embedded in Tissue Tek (Sakura Finetek, Zouterwoude, The Netherlands), and coronal cryo-sections (30 µm) were prepared.

For immunohistochemical staining, sections were incubated in 0.3% (v/v) H₂O₂ for 30 min at room temperature before unspecific binding sites were blocked for 1 h at room temperature in blocking solution [0.75% (v/v) Triton X-100, 5% (v/v) normal goat serum (NGS), 5% (v/v) normal donkey serum (NDS), in PBS]. Subsequently, slices were incubated at 4°C for 3 days with primary antibodies (active Casp-3, 1:50, ab2302, Abcam, Milton, UK; GFAP, 1:500, MAB3402, Millipore, Hohenbrunn, Germany) diluted in incubating solution [0.75% (v/v) Triton X-100, 0.5% (v/v) NGS, 0.5% (v/v) NDS, in PBS]. After washing with PBS, samples were incubated at room temperature for 4 h with secondary antibodies (gt-α-msA647, 1:50, A-21235, Life Technologies, Carlsbad, CA, USA; gt-α-rbA555, 1:500, A-21428, Life Technologies) diluted in incubating solution. For DAPI staining, NucBlue Reagent (Life Technologies) was used according to the supplier's protocol. Slices were washed, transferred onto slides (SuperFrost Plus, Menzel, Braunschweig, Germany) and, finally, a coverslip was fixed on the samples with Aqua-Poly/Mount (Polysciences, Eppelheim, Germany). Fluorescent images were obtained with an inverted confocal microscope (TCS SP8, Leica, Wetzlar, Germany).

We applied recombinant adeno-associated virus serotype 9 (rAAV9) vectors for injection into the murine hippocampus. Viral vectors were designed to mediate expression of shRNA constructs targeting independent genes, i.e. hcn and luc. Because of their low immunogenicity, rAAV vectors are generally considered to be an optimal tool for genetic manipulation in vivo.

We designed viral constructs to express two distinct shRNA-encoding fragments as well as eGFP for the identification of infected cells (Fig. 1). Prior to stereotaxic injection, viral constructs were tested for knockdown efficiency and specificity of endogenous hcn1 and hcn2 expression in primary hippocampal neurons by qPCR. We assessed cell viability and morphology in primary cultures by immunofluorescence microscopy of neurons co-expressing shRNA and eGFP in comparison either to neurons transduced with rAAV vectors mediating only eGFP expression or to non-transduced control cultures. No detrimental effects on the health of neurons were observed in these cultures (data not shown).

After establishing the functionality of shRNAs in vitro, rAAV9 vectors were injected into the dorsal hippocampus. Animals were killed at 3–8 weeks post-infection. We observed widespread eGFP fluorescence in hippocampal tissue after a single stereotaxic injection of rAAV9 constructs (Fig. 2). Expression of eGFP was driven by the neuron-specific CaMKIIα promoter, whereas shRNAs were expressed under the control of the constitutive hU6 promoter. Because of vector design, transduced neurons can be identified on the basis of eGFP fluorescence, whereas transduced non-neuronal cells do not show eGFP fluorescence, but could still express shRNAs.

Some animals injected with shRNA-encoding rAAVs displayed neuronal degeneration at 3 weeks post-infection (Fig. 2D). In contrast, eGFP expression alone did not induce an adverse tissue response (Fig. 2B), confirming that the degeneration was not due to eGFP expression per se. Tissue degeneration advanced with time and was even stronger in animals injected with luc-targeting control constructs than with hcn-targeting constructs.

In order to further assess the observed tissue degeneration, we examined the expression of potential cellular stress markers (Fig. 3), including Casp-3 and GFAP. As all of the applied rAAV constructs encoded eGFP as a marker, transduction of the dorsal hippocampus was monitored based on eGFP fluorescence. Strong immunoreactivity of Casp-3 and GFAP was observed in the dorsal hippocampus at 3 weeks post-infection even though tissue degeneration was not yet apparent. Neither Casp-3 nor GFAP was detected in the ventral hippocampus at this time point. At 4 weeks post-infection, degeneration in the dorsal hippocampus became apparent and immunoreactivity of Casp-3 and GFAP was also detected in more ventral regions of the hippocampal formation. Notably, detection of Casp-3 and GFAP was always in concordance with eGFP expression.

RNAi-based strategies frequently use viral vectors for specific manipulation of gene expression at defined tissue sites in the rodent brain. Previous studies have employed shRNA constructs without apparent detrimental effects on cell viability in the hippocampus (Mosser et al., 2015; Omata et al., 2011; Kim et al., 2012). In this study, we applied virus titers within the range chosen by most rAAV-based RNAi studies in the CNS, i.e. between 10⁹ and 10¹¹ total virus particles. We examined three different eGFP-encoding rAAV9 constructs. Viral vectors mediated expression of: (1) shRNAs targeting endogenous hcn genes, (2) shRNAs targeting the luc gene, which is not endogenously expressed in mammals and is used widely as a control in similar studies, and (3) a negative control without a shRNA-encoding sequence. Surprisingly, even though no adverse effects of the shRNA-encoding constructs were observed in primary hippocampal neurons cultivated in vitro, rAAV9-mediated shRNA expression induced strong cellular degeneration in the hippocampus of mice (see Fig. 2). The detrimental effects were independent of the encoded shRNA sequence as they were observed for constructs targeting both endogenous and non-endogenous transcripts (i.e. hcn and luc), but were not apparent in animals treated with rAAVs encoding only eGFP.

Cytotoxic effects of shRNAs have previously been shown to be caused by overloading of the endogenous miRNA machinery (Börner et al., 2013; Grimm et al., 2006, 2010; McBride et al., 2008; Yi et al., 2005). These observations were corroborated by the pronounced shRNA dose dependency of cytotoxicity (Grimm, 2011). However, as the total number of virus particles may vary depending on the applied construct and the treated CNS region, potential effects on the tissue must be assessed for each construct, individually. As AAV9 is known to transduce neurons as well as astrocytes (Aschauer et al., 2013), we cannot rule out that constitutive shRNA expression in astrocytes might be an additional factor in the observed hippocampal degeneration.

So far, reports examining shRNA-induced cytotoxicity have not consistently described the expression of molecular markers related to tissue degeneration (Bauer et al., 2009; Hutson et al., 2014; Martin et al., 2011; McBride et al., 2008). To address this issue in the context of the observed hippocampal degeneration, we assessed expression of Casp-3, a marker of apoptosis (Ashkenazi and Salvesen, 2014), and GFAP, a marker of astroglial activation (Sofroniew, 2009). Expression of both molecular markers strongly correlated with rAAV9-mediated shRNA expression.

Our results emphasize that detrimental effects of shRNA application in vivo might not always be readily identified in cell culture. We suggest that the design of shRNA-based studies should involve careful consideration of several parameters, such as shRNA dose (Grimm, 2011), promoter choice (Lebbink et al., 2011; Sun et al., 2013), AAV serotype (Ehlert et al., 2010) and construct backbone (Boudreau et al., 2009; Han et al., 2011; McBride et al., 2008). These factors should subsequently be tested with respect to titer, construct and specificity in vivo. Furthermore, RNAi-based studies should take into account that the detrimental effects observed here for standard C57BL/6J mice may differ depending on the employed mouse strains. However, the identification of reliable markers, such as increased Casp-3 and GFAP expression, for monitoring shRNA-induced tissue stress even before the onset of degeneration is a requirement for validating behavioral experiments that are based on RNAi strategies.

We gratefully acknowledge the assistance of Dr H. Büning with establishing the rAAV cell culture. We thank Dr D. Kaschuba for her help with the identification of hcn-targeting shRNA constructs.