Neuron type-specific expression of a mutant KRAS impairs hippocampal-dependent learning and memory

KRAS mutations are associated with rare cases of neurodevelopmental disorders that can cause intellectual disabilities. Previous studies showed that mice expressing a mutant KRAS have impaired the development and function of GABAergic inhibitory neurons, which may contribute to behavioural deficits in the mutant mice. However, the underlying cellular mechanisms and the role of excitatory neurons in these behavioural deficits in adults are not fully understood. Herein, we report that neuron type-specific expression of a constitutively active mutant KRASG12V in either excitatory or inhibitory neurons resulted in spatial memory deficits in adult mice. In inhibitory neurons, KRASG12V induced ERK activation and enhanced GABAergic synaptic transmission. Expressing KRASG12V in inhibitory neurons also impaired long-term potentiation in the hippocampal Shaffer-collateral pathway, which could be rescued by picrotoxin treatment. In contrast, KRASG12V induced ERK activation and neuronal cell death in excitatory neurons, which might have contributed to the severe behavioural deficits. Our results showed that both excitatory and inhibitory neurons are involved in mutant KRAS-associated learning deficits in adults via distinct cellular mechanisms.

Inhibitory synaptic transmission is increased in the inhibitory neuron-specific KRAS G12V expressing mice. To assess whether the overexpression of KRAS G12V in inhibitory neurons affects their electrophysiological properties, we measured spontaneous inhibitory postsynaptic currents (sIPSC) in the hippocampal pyramidal neurons in vGAT-Cre::KRAS G12V mice using whole-cell patch-clamp recordings. We found that the sIPSC frequency was significantly higher in vGAT-Cre::KRAS G12V mice than in vGAT-Cre::EYFP mice (Fig. 2a). Ectopic expression of KRAS G12V in vGAT + inhibitory neurons did not affect the sIPSC amplitude (Fig. 2a). The frequency and amplitude of the spontaneous excitatory postsynaptic currents (sEPSC) in the KRAS G12V group were comparable to those of the EYFP group (Fig. 2b). These data show that the activation of KRAS-ERK pathway enhances GABAergic synaptic transmission. Next, we examined whether the changes in GABAergic synaptic transmission affects synaptic plasticity in the hippocampus. We recorded field excitatory postsynaptic potentials (fEPSPs) in the hippocampal CA3-CA1 Schaffer-collateral pathway. We found that the inhibitory neuron-specific expression of KRAS G12V significantly decreased the input-output relationship of fEPSPs (Fig. 2c). Both the initial slope and amplitude of fEPSP were decreased in vGAT-Cre::KRAS G12V (Fig. 2c). The paired-pulse facilitation ratio was significantly decreased at 50 ms of inter-stimuli interval in vGAT-Cre::KRAS G12V comparison to vGAT-Cre::EYFP (Fig. 2d). Long-term potentiation (LTP) is considered as a cellular mechanism for memory and LTP impairments have been frequently observed in mutant mice with learning and memory deficits including RASopathy models 16,23 . However, a weak theta-burst stimulation (TBS; 4 bursts of 4 100 Hz pulses, 5 Hz inter-burst interval)-induced LTP was normal in the hippocampus of vGAT-Cre::KRAS G12V mice (Fig. 2e). When LTP was induced by the stronger stimulation protocol (10 bursts of 4 100 Hz pulses, repeated 4 times with 10 s interval), we found that LTP is significantly reduced in vGAT-Cre::KRAS G12V in comparison with that in vGAT-Cre::EYFP (Fig. 2f). To assess whether the LTP deficit is caused by increased inhibitory function, we induced LTP with the strong protocol in the presence of a GABA A receptor antagonist Taken together, our results suggest that inhibitory neuron-specific expression of KRAS G12V induces deficits in synaptic plasticity via increased basal inhibitory synaptic transmission.

Ectopic expression of KRAS G12V in the hippocampal excitatory neurons impairs spatial learning and memory in mice.
In adult mice, Kras is expressed not only in inhibitory neurons but also in excitatory neurons 19 . To investigate the effects of ectopic KRAS G12V expression in the excitatory neurons on learning and memory, we injected AAV-KRAS G12V -HA into the dorsal hippocampus of αCaMKII-Cre mice (Fig. 3a), thereby allowing the selective expression of KRAS G12V in the excitatory neurons 24 . We then subjected the mice to the MWM test. αCaMKII-Cre::KRAS G12V mice showed a significantly increased escape latency during the training sessions of the MWM test, suggesting that KRAS G12V expression in the excitatory neurons severely impaired spatial learning (Fig. 3b). In probe trials, αCaMKII-Cre::KRAS G12V mice failed to identify the target quadrant, whereas the control αCaMKII-Cre::EYFP mice spent a significantly longer time in the target quadrant, suggesting that KRAS G12V expression in the adult hippocampal excitatory neurons also impairs spatial memory (Fig. 3c). Moreover, αCaMKII-Cre::KRAS G12V mice swam farther from the platform location ( Supplementary  Fig. S3a) and crossed the platform fewer times than the EYFP-expressing control mice ( Supplementary Fig. S3b). Of note, αCaMKII-Cre::KRAS G12V mice swam significantly slower than αCaMKII-Cre::EYFP mice during the probe trials and showed an increased latency to find the platform in the initial trials of the visible platform version of the MWM. However, they performed well in the later trials, suggesting that expressing KRAS G12V in hippocampal inhibitory neurons may also affect locomotive behaviours (Supplementary Fig. S3c and d). However, even when we analysed the path length to the platform during the training sessions in the hidden platform version of the MWM, αCaMKII-Cre::KRAS G12V mice took longer path to the platform than αCaMKII-Cre::EYFP mice, demonstrating that the learning deficits in the αCaMKII-Cre::KRAS G12V mice cannot be attributed to the altered locomotive behaviours ( Supplementary Fig. S3e). Furthermore, αCaMKII-Cre::KRAS G12V mice showed significantly lower freezing time compared to αCaMKII-Cre::EYFP mice in contextual test, but not in auditory test in the classical fear conditioning (Fig. 3d,e), showing that expressing KRAS G12V in excitatory neurons impairs hippocampus-dependent memory without affecting amygdala-dependent memory. Interestingly, αCaMKII-Cre::KRAS G12V mice showed hyperactivity and avoided the center zone in the open field test (Supplementary Fig. S3f and g), suggesting that the behavioural deficits in αCaMKII-Cre::KRAS G12V mice are not limited to learning and memory.

Ectopic expression of KRAS G12V in the excitatory neurons increases ERK activity and induces neuronal cell death.
We observed that ectopic expression of HA-tagged KRAS G12V in the hippocampal CA1 region of αCaMKII-Cre mice caused hyperactivation of RAS-ERK signalling as accessed by pERK immunostaining (Fig. 3f,g). Interestingly, we also observed that the hippocampal CA1 region showed morphological alterations such as the decrease in cell density and the expansion of the pyramidal layer (Fig. 3f). To visualize the morphological abnormalities, we performed Nissl staining and immunohistochemistry on hippocampal slices from EYFP or KRAS G12V expressing mice ( Supplementary Fig. S4). The pyramidal cell layer was distended, but the dentate gyrus, where the mutant protein was not expressed, appeared normal ( Supplementary Fig. S4). This morphological alteration might have been due to cell death. To test whether ectopic KRAS G12V expression induced neuronal cell death, we examined the expression of cleaved caspase-3, which is an apoptosis-dependent cell death marker 25,26 . The number of cleaved caspase-3 positive neurons was significantly increased in the  www.nature.com/scientificreports/ αCaMKII-Cre::KRAS G12V mice compared to αCaMKII-Cre::EYFP mice (Fig. 4a,b). Moreover, we observed that ectopic KRAS G12V expression did not induce apoptosis in inhibitory neurons (Fig. 4c,d).

Discussion
Herein, we report that ectopic expression of a gain-of-function KRAS mutation in either excitatory or in inhibitory neurons in the adult hippocampus increases ERK activation and impairs spatial learning and memory in mice. In this study, we used a strong G12V KRAS mutation to mimic severe cases of RASopathies caused by strong KRAS mutations such as G12S to investigate the cellular mechanisms of cognitive deficits in these RASopathies. www.nature.com/scientificreports/ Cell type specific roles of RAS signalling in learning and memory. It has been shown that the RAS proteins are critically involved in learning and memory and synaptic plasticity 16,27 . Hyperactivation as well as hypoactivation of RAS-ERK signalling impairs learning and memory and synaptic plasticity 16,[27][28][29] . More interestingly, mutations in the essential genes of RAS-ERK signalling cascade have cell type-specific effects on the downstream signalling activation as well as synaptic functions. Nf1 haploinsufficiency affects only GABAergic synaptic transmission in the cortex and the hippocampus 30,31 . Consistently, inhibitory neuron-specific deletion of Nf1 recapitulated learning deficits of Nf1 +/mice 30 . Cell type-specificity has also been reported for excitatory neurons. SHP2 D61G/+ mice showed increased level of ERK activation only in excitatory neurons and enhanced glutamatergic synaptic transmission but no changes in inhibitory synaptic functions 32 . Ectopic expression of SHP2 D61G selectively in the excitatory but not in inhibitory neurons increased ERK activation and impaired spatial learning and LTP 19 . One explanation for these cell type-specific effects of RAS signalling is that the expression profiles of RAS-ERK signalling genes are significantly different between hippocampal excitatory and inhibitory neurons 19 . The NF1 transcript is significantly enriched in inhibitory neurons compared to excitatory neurons in both mice as well as in humans, which may account for its critical roles in regulating inhibitory synaptic transmission 19,33 . Similarly, Kras mRNA expression was shown to be significantly higher in inhibitory neurons than excitatory neurons in the adult mouse hippocampus, suggesting that Kras mutations may have a more severe impact on inhibitory neurons 19 . Consistent with this hypothesis, a heterozygous knock-in mouse expressing KRAS G12V under the control of a synapsin promoter (Syn-cre; KRAS G12V mice) showed increased inhibitory synaptic functions and deficits in LTP and spatial memory 18 . Moreover, inhibitory neuron-specific expression of KRAS G12V also increased the GABAergic synaptic area and impaired the memory functions, demonstrating the critical role of KRAS G12V in inhibitory neurons 18 . Consistently, our results also show that KRAS is involved in regulating GABA release in inhibitory neurons. Interestingly, we found that KRAS G12V induces cell death only in excitatory neurons, and not in inhibitory neurons, thereby demonstrating that KRAS also plays cell type specific roles in the nervous system.

Role of KRAS in inhibitory neurons. Papale et al. showed that ERK hyperactivation is only observed in
the early phases of postnatal development (P10 and P21) in Syn-cre;KRAS G12V mice 18 . They also reported that reducing GABAergic inhibition but not suppressing ERK hyperactivation rescued the LTP and memory deficits in the adult mutants, suggesting that increased RAS-ERK signalling mainly accounts for the developmental defects such as increased inhibitory synaptogenesis 18 . However, the effects of cell-type specific activation of ERK signalling in the adult hippocampus have not been fully examined. Moreover, it remained unclear whether KRAS G12V in mature neurons contributes to the behavioural and physiological deficits in Syn-cre;KRAS G12V mice in which KRAS G12V is expressed in all types of neurons. In this study, we found that KRAS G12V increases ERK activation in the adult hippocampal inhibitory neurons, which is sufficient to cause deficits in synaptic plasticity and spatial memory. Our results demonstrate that KRAS G12V -induced increase in ERK activity in inhibitory neurons is at least partially responsible for the behavioural deficits in the adult mice, which can be exploited for developing alternative treatment strategies for targeting RASopathies. The RAS-ERK signalling cascade regulates GABA release by phosphorylating synapsin at the inhibitory synapse 30 . Consistently, we also observed that expressing KRAS G12V in inhibitory neurons increases ERK activation and sIPSC frequency. However, a detailed molecular mechanism how KRAS G12V increases the spontaneous GABA release remains to be investigated. We also found that the input-output curve of fEPSP at the CA3-CA1 synapse is significantly suppressed by expressing KRAS G12V in inhibitory neurons, which might be caused by the constitutively increased inhibitory tone. It has been shown that dendritic inhibition is effective at shunting postsynaptic depolarization 34 .
It has been shown that increased GABAergic synaptic transmission can impair LTP 18,30,35 . Interestingly, LTP was not impaired in vGAT-Cre::KRAS G12V mice when we used a weak stimulation protocol for LTP induction which was also used in Syn-cre;KRAS G12V mice exhibiting LTP deficits 18 . Differences in mouse models such as the generation methods (knock-in versus viral overexpression, synapsin promotor versus vGAT promoter) may account for this discrepancy. However, when we used the stronger stimulation protocol (4 episodes of 10 TBS) which should activate more inhibitory synapses 35 , we were able to reveal significant LTP deficits in vGAT-Cre::KRAS G12V mice. Picrotoxin treatment rescued the LTP deficit, suggesting that the increased inhibition is responsible for the LTP deficits. Taken together, our results suggest that KRAS in adult vGAT + neurons is critically involved in regulating GABA release, and that its dysregulation results in deficits in synaptic plasticity and memory.
Role of KRAS in excitatory neurons. Next, we assessed whether excitatory neurons also contribute to the phenotypes in KRAS mutations-associated neurodevelopmental disorders. We observed that, similar to inhibitory neurons, the ectopic expression of KRAS G12V specifically in excitatory neurons also resulted in severe spatial learning and memory deficits in adult mice. Interestingly, KRAS G12V expression induced neuronal cell death selectively in excitatory neurons, but not in inhibitory neurons. Thus, neuronal apoptosis in KRAS G12V expressing mice prevented us from analysing the electrophysiological properties in these mice. Moreover, KRAS G12V expression induced histological changes such as an expanded pyramidal cell layer. Not surprisingly, αCaMKII-Cre::KRAS G12V mice showed complex behavioural phenotypes such as increased locomotive activity and anxiety-like behaviours in the open field test, which may also contribute to the learning and memory phenotypes. Consistent with our findings, it has been shown that nestin-derived KRAS G12V expression causes apoptosis in neural progenitor cells, leading to severe brain oedema in zebrafish 36 . Although RAS-ERK signalling is well known for its anti-apoptotic or pro-survival functions, it also has pro-apoptotic functions 37 . For example, in several conditions such as focal cerebral ischemia, DNA damaging stimuli and oxidative toxicity promote cell death Scientific Reports | (2020) 10:17730 | https://doi.org/10.1038/s41598-020-74610-y www.nature.com/scientificreports/ through ERK signalling 37 . Moreover, it was shown that ERK activation is required for caspase-3 activation 38 . Activation of ERK signalling may elicit either apoptosis or cell survival depending on conditions, the mechanisms of which are not fully known. In addition, the mechanisms underlying KRAS-induced apoptosis remain unknown. In addition, it remains to be investigated how KRAS G12V induces apoptosis only in excitatory neurons, but not in inhibitory neurons. As previously shown, since RAS signalling networks are different between excitatory and inhibitory neurons, we speculate that KRAS-ERK signalling pathway is not coupled to pro-apoptotic signalling in inhibitory neurons. Previous studies have shown that mutations in genes involved in the RAS signalling pathway can alter the composition of cell types in both the developing and adult brain 16,39,40 . Changes in the ratio of excitatory and inhibitory neurons can subsequently distort the excitation-inhibition balance, which is critical for normal cognitive functions, including learning and memory 41,42 . Excitatory neuron-specific cell death in adults, which contribute to behavioural deficits, might have been overlooked and may also contribute to the increased inhibition to excitation ratio which has been reported in other RASopathy models 16,18,30,32 . However, we cannot completely exclude the possibility that the disrupted cellular organization in CA1, including the enlarged intercellular space in CA regions, might also have affected the synaptic connectivity in the hippocampus, thereby contributing to behavioural deficits.
Taken together, our study's results showed that a gain-of-function KRAS mutation affects the function and survival of inhibitory and excitatory neurons, respectively. Further, this may have synergistic effects on cognitive deficits in RASopathies (Supplementary Fig. S5). Finally, our findings may contribute to better understanding the physiological roles of KRAS in adult neurons.

Methods
Mice. αCaMKII-Cre (JAX 005,359) and vGAT-IRES-Cre (JAX 016,962) mice were maintained by breeding with wild-type C57Bl/6 J mice at the SNU Specific Pathogen Free centre. Animals were group-housed (four mice per cage) on a 12-h light/dark cycle in the vivarium at SNU. Both female and male mice (2-7 months old) were used for all experiments. All studies were approved by the Seoul National University Institutional Animal Care and Use Committee. All experiments were performed in accordance with the institutional guidelines and regulations.

Stereotaxic viral injection.
Stereotaxic surgery was performed as previously described 19  Morris water maze (MWM) test. The MWM test was performed as previously described 19 . Briefly, mice were handled for 2 min at the same time every day for seven consecutive days before the test. The maze consisted of a non-transparent cylindrical tank (diameter, 120 cm) in a room with visual cues, which are the animal's navigational references for locating the platform. The tank was filled with water (21-22 °C) and painted white. The tank was divided into four invisible quadrants (target quadrant, T; opposite quadrant, O; right quadrant, R; left quadrant, L.) and a platform, which was submerged 1 cm under the surface of the water, was placed at the centre of the target quadrant. Before the first trial on training day 1, each mouse was placed onto the platform for 30 s. On training days, mice were placed at the start position, chosen randomly for each trial. Mice were trained with six trials per day. Probe tests were performed under the same conditions as the training trials, except that the platform was absent in the test session. In the probe tests, the mice were tracked for 1 min using a tracking program (EthoVision 11.5; Nodulus). For vGAT-Cre::KRAS G12V and vGAT-Cre::EYFP mice, the probe test was performed on day 7 after the training. For CaMKII-Cre::KRAS G12V and CaMKII-Cre::EYFP mice, the probe test was performed 5 days after the training. The visible platform-version of MWM was performed after the hidden- Object-place recognition (OPR) test. OPR test was performed as described previously 19 . Mice were handled for 5 min for four consecutive days and habituated in a cube-shaped acrylic box (32 cm by 32 cm by 32 cm) for 15 min for another 2 days before performing the training and test. In the training session, mice were placed in the box containing two identical 100 ml glass bottles and were allowed to explore the objects for 10 min. In the test session, 24 h after training, mice were placed in the same box containing one object that stayed in the same location and the other object relocated to a new position. All locations for the objects were counterbalanced among groups, and objects were cleaned between trials. Sessions were recorded and later analyzed manually. Experimenters were blinded to the type of injected viral vectors.
Classical fear conditioning. Classical fear conditioning test was performed as previously described 43 .
Mice were placed in the conditioning chamber for 2 min before the delivery of three times of paired stimuli containing a tone (2800 Hz, 85 dB, 30 s) and a co-terminating electric foot shock (2 s, 0.4 mA). Contextual fear memory was assessed 24 h after training, in which the percentage of time mice spent freezing was measured in the same chamber where they were previously shocked. In cued fear memory test performed 1 h or 1 day after the contextual memory test, mice were placed in a novel context for 2 min followed by the conditioned tone for 3 min. The percentage of freezing time was automatically scored by Freeze Frame software (ActiMetrics, IL., USA).
Immunohistochemistry. Immunohistochemistry was performed as described previously 19

Electrophysiology
Field recording. Field EPSP recordings were performed as previously described 32 . Briefly, mouse brains were sliced into sagittal Sects. (400 μm thickness) using a vibratome (Campden Instruments, Oxford, UK) in ice-chilled slicing solution (2.5 mM KCl, 30 mM NaHCO 3 , 93 mM NMDG, 20 mM HEPES, 2 mM thiourea, 1.25 mM NaH 2 PO 4 , 10 mM MgSO 4 , 25 mM d-glucose, 5 mM sodium ascorbate, 3 mM sodium pyruvate, and 0.5 mM CaCl 2 , oxygenated with 5% CO 2 and 95% O 2 ). Slices were allowed to recover in the slicing solution at 32 °C for 15 min. Slices were incubated in ACSF (3.5 mM KCl, 1.25 mM NaH 2 PO 4 , 1.3 mM MgSO 4 , 120 mM NaCl, 2.5 mM CaCl 2 , 10 mM glucose, 20 mM NaHCO 3 , oxygenated at room temperature for at least 30 min before recording). Slices were translocated into a recording chamber and fEPSPs were recorded from the Schaffer collaterals in the CA3-CA1 pathway. A stimulation intensity of 30-40% of the maximum response was chosen for these studies. All the electrophysiological experiments were performed as previously described 32 . The input-output relationship was determined by measuring the fEPSP slope at stimulation intensities (0-100 μA). The paired-pulse facilitation ratio was analysed over increasing time intervals (10,25,50,100,200, and 400 ms). LTP was induced by either a weak theta-burst stimulation (TBS) protocol (4 bursts of stimuli delivered at 5 Hz; each burst contains four pulses at 100 Hz) or strong TBS protocol (4 episodes of TBS with 10 s interval; each episode consists of 10 TBS). Data were analysed using WinLTP software (WinLTP Ltd., Bristol, UK).
Whole cell patch clamp recording. The whole-cell patch was performed as described previously 44 . Briefly, transverse hippocampal slices (300 μm) were prepared using a vibratome (Leica, VT1200S) in an ice-chilled slicing solution (2.5 mM KCl, 30 mM NaHCO 3 , 93 mM NMDG, 20 mM HEPES, 2 mM thiourea, 1.25 mM NaH 2 PO 4 , 10 mM MgSO 4 , 25 mM d-glucose, 5 mM sodium ascorbate, 3 mM sodium pyruvate, and 0.5 mM CaCl 2 , saturated with 5% CO 2 and 95% O 2 ). Slices were allowed to recover in the slicing solution at 32 °C for 30 min. The slices were transferred into an incubation chamber filled with ACSF solution saturated with 5% CO 2 and 95% O 2 and maintained for at least 1 h before recordings were made. The whole-cell patch-clamp recording was performed at 30 °C during continuous perfusion at 4 ml/min with ACSF. All recordings were performed using EPC 10 (HEKA Elektronik) with a sampling frequency of 20 kHz and the signals were filtered at 10 kHz. Patch pipettes (3-4 MΩ) were borosilicate glass filled with the internal solution (135 mM CsMS; 10 mM CsCl; 10 mM HEPES; 0.2 mM EGTA; 4 mM Mg-ATP; 0.4 mM Na2-GTP; 0.5 mM spermine; ~ 295 mOsm; pH 7.3 adjusted with CsOH). Experiments were only accepted for analysis if series resistance values were < 25 megaohms (± 20%) throughout the recordings. Cells were clamped at a holding potential of − 70 mV to measure the spontaneous excitatory postsynaptic current (sEPSC). Cells were clamped at a holding potential of + 10 mV to measure the spontaneous inhibitory postsynaptic current (sIPSC).