Less means more: The magnitude of synaptic plasticity along the hippocampal dorso‐ventral axis is inversely related to the expression levels of plasticity‐related neurotransmitter receptors

Abstract The dorsoventral axis of the hippocampus exhibits functional differentiations with regard to (spatial Vs emotional) learning and information retention (rapid encoding Vs long‐term storage), as well as its sensitivity to neuromodulation and information received from extrahippocampal structures. The mechanisms that underlie these differentiations remain unclear. Here, we explored neurotransmitter receptor expression along the dorsoventral hippocampal axis and compared hippocampal synaptic plasticity in the CA1 region of the dorsal (DH), intermediate (IH) and ventral hippocampi (VH). We observed a very distinct gradient of expression of the N‐methyl‐D‐aspartate receptor GluN2B subunit in the Stratum radiatum (DH< IH< VH). A similar distribution gradient (DH< IH< VH) was evident in the hippocampus for GluN1, the metabotropic glutamate receptors mGlu1 and mGlu2/3, GABAB and the dopamine‐D1 receptor. GABAA exhibited the opposite expression relationship (DH > IH > VH). Neurotransmitter release probability was lowest in DH. Surprisingly, identical afferent stimulation conditions resulted in hippocampal synaptic plasticity that was the most robust in the DH, compared with IH and VH. These data suggest that differences in hippocampal information processing and synaptic plasticity along the dorsoventral axis may relate to specific differences in the expression of plasticity‐related neurotransmitter receptors. This gradient may support the fine‐tuning and specificity of hippocampal synaptic encoding.

Moreover, a functional separation of this kind along the hippocampal longitudinal axis is likely to depend on differentiated preferences for input information and information processing within the intrinsic neuronal circuits of the dorsal, intermediate, and ventral subdivisions.
This may be reflected in the form of differences in synaptic plasticity, as the means through which synaptic information encoding takes place.
Correspondingly, it has been reported that the VH expresses much weaker synaptic potentiation compared with its dorsal counterpart (Maggio and Segal, 2007a;Maggio, Shavit Stein, & Segal, 2015;Maruki, Izaki, Nomura, & Yamauchi, 2001;Papatheodoropoulos and Kostopoulos, 2000). In contrast, other studies have reported that the ventral hippocampal pole expresses quite strong, or even equivalent synaptic potentiation compared to potentiation elicited in the dorsal CA1 hippocampus (Kouvaros and Papatheodoropoulos, 2016). However, in case of these former reports, LTP was evoked with high-frequency stimulation, whereas Kouvaros and Papatheodoropoulos (2016) used theta burst stimulation (TBS) protocols with varying numbers of bursts-from 1 to 8. Thus, there seems to be little consistency with regard to protocols used and LTP responses obtained. Similarly, the profile of longterm synaptic depression in the ventral pole was reported to be either equivalent to, or of greater magnitude than, synaptic depression evoked in the dorsal CA1 (Izaki, Takita, & Nomura, 2000;Maggio and Segal, 2009a). These differences may have arisen due to differences in the hippocampal slice preparation, or afferent frequencies used to elicit synaptic plasticity. Another point that remains unclear is whether the intermediate CA1 region is able to express long-term forms of synaptic plasticity, such as long-term potentiation (LTP) or long-term depression (LTD). To our knowledge, synaptic plasticity in the IH of the rat has only been examined in the dentate gyrus (Kenney and Manahan-Vaughan, 2013a,b). However, in mouse hippocampal slices, the intermediate CA1 region was shown to produce LTP of intermediate magnitude between the dorsal and ventral CA1 responses that was sustained/monitored for 60 minutes (Milior et al., 2016).
Knowledge of this kind is essential for both the interpretation and understanding of how synaptic plasticity properties may be differentiated along the hippocampal longitudinal axis.
The aim of this study was, firstly, to characterize and compare the expression and distribution of plasticity-related proteins in the CA1 region of the dorsal, intermediate and ventral subdivisions of the hippocampus. We focused on examining subunits of the NMDAR (GluN1, GluN2A, and GluN2B), groups I and II mGlu receptors (mGlu1, mGlu5, and mGlu2/3), GABAergic receptors, and dopaminergic receptors. Our second aim was to investigate physiological properties of the CA1 neurons along the hippocampal longitudinal axis, as well as to compare the ability of the Schaffer collateral-CA1 synapses to express long-term synaptic plasticity. We identified an expression profile for glutamatergic, GABAergic, and dopaminergic receptors that was distinct for the dorsal, intermediate, and ventral hippocampal parts. This was associated with differences in physiological properties of the CA1 neurons, along with differences in the ability of the dorsal, intermediate, and ventral hippocampal parts to express both LTP and LTD. Taken together, these data suggest that the dorsal, intermediate, and ventral hippocampal parts exhibit physiologically distinct properties, and that these distinctions are enabled at least in part, by the very different expression profiles of plasticity-related proteins exhibited along the dorsoventral hippocampal axis. These differences may serve to explain the functional heterogeneity that is attributed to the dorsoventral axis of the hippocampus (Fanselow & Dong, 2010;Strange et al., 2014).

| Animals
All experiments were conducted using 6-10-week-old male Wistar rats (Charles River Laboratories, Sulzfeld, Germany). Animals were housed in custom-made climatised and ventilated holding cupboards, in an animal-housing room with a controlled 12-h light/dark cycle. No female rats were housed in the room. Animals had free access to food and water. The study was carried out in accordance with the European Communities Council Directive of September 22, 2010 (2010/63/EU) for care of laboratory animals.

| Immunohistochemistry
For immohistochemical analysis, animals were euthanized with sodium pentobarbital and transcardially perfused with cold Ringer's solution 1 heparin (0.2%), followed by 4% paraformaldehyde (PFA) in phosphate buffered saline (PBS, 0.025 M). Brains were then removed, fixed in 4% PFA for 24 h-, and cryoprotected in 30% sucrose in 0.1 M PBS for at least 3 days. Serial 30-mm thick horizontal sections were collected using a freezing microtome. For each animal (N 5 10), three DUBOVYK AND MANAHAN-VAUGHAN | 137 horizontal sections from the most dorsal (between 3.6 and 4.1 mm posterior to bregma), middle intermediate (around 5.6 mm posterior to bregma) and most ventral hippocampal parts (between 7.1 and 7.6 mm posterior to bregma) were simultaneously used for immunohistochemical staining (Figure 1 and Supporting Information Figure S1). Freefloating brain sections were pretreated in 0.3% H 2 O 2 in PBS for 20 min, rinsed in PBS and then incubated with blocking solution containing 10% normal serum 1 20% avidin in PBS with 0.2% Triton X-100 (PBS-   . Given that images were acquired with a red, green, blue camera, the "Color Deconvolution" plugin in ImageJ was used to deconvolve the color information and to convert images to eight-bit format, thus, increasing the dynamic range of color representation (Jacqui Ross, 2014). As a next step, the background staining was subtracted from each image. In the dorsal hippo- were averaged and then this averaged value was subtracted from corresponding images. Finally, R software was used to scale data from several independent stainings/plates using generalized residual sum of squares algorithm to account for batch effects of staining intensities (Kreutz et al., 2007;von der Heyde et al., 2014).

| Immunoblotting
Two protein biochemical methods were used as specificity controls for immunohistochemical experiments (Supporting Information Figure S2): 1. direct immunoblot analysis from whole tissue lysate or, 2. immunoprecipitation that was followed by immunoblotting.
For immunoblotting experiments, brains were rapidly removed followed by whole hippocampus dissection. The tissue were then homogenized in 20 mM Tris-HCl buffer (pH 7.4) containing 10% sucrose to 1:20,000 dilution ranges for 90 min at room temperature. Protein bands were visualized using an enhanced chemiluminescence reagent (ECL), Pierce ECL Plus, or ECL Prime, on X-ray films or CCD camera.
For immunoprecipitation experiments, a tissue lysate in a volume of 200 mg was filled up to 400 mL of total volume with sample buffer.
25 mL of 50% Protein A Sepharose (PAS) beads and 4 mL of primary antibodies (anti-NMDAE1 or anti-NMDAE2) were added. Immunocomplexes were captured through overnight incubation at 4 8C. PAS beads were then briefly centrifuged (20-30 s) and rinsed in sample buffer.
The procedure was repeated three times. After the last rinsing, 25-30 mL of liquid were left on top of the Sepharose beads and the same amount of 23 Laemmli buffer was added. From here on, immunoprecipitation was followed by immunoblotting as described earlier.

| In vitro electrophysiology
Brains were dissected in ice-cold (1-4 8C), oxygenated artificial cerebrospinal fluid (aCSF) containing (in mM: 124 NaCl; 4.9 KCl; 1.2 NaH 2 PO 4 ; 1.3 MgSO 4 ; 2.5 CaCl 2 ; 25.6 NaHCO 3 ; and 10 D-glucose; pH 7.4). The two hippocampi were isolated and then sectioned into 400-mm thick slices using a vibratome (VT 1000S, Leica, Nussloch, Germany). Specifically, transverse slices of the longitudinal axis of the hippocampus were prepared as shown on Figure 1. Here, slices from the dorsal, intermediate and ventral hippocampal subdivisions were used. Slices were incubated for 15 min at 35 8C, and then placed on a nylon net in separate submerged recording chambers for at least 1 h prior to any recordings.
Slices were continuously perfused at a constant flow rate of 2 mL/min with oxygenated aCSF at 32-33 8C. Field recordings were made with a metal electrode (platinum/tungsten core, impedance: 0.5 MX; Thomas Recording, Gießen, Germany) positioned in the sr of the CA1 region.
Stimulation was delivered through a bipolar electrode (Fredrick Haer, Bowdowinham, ME) placed in the Schaffer collaterals. Test-pulse stimuli of 0.2 ms duration were applied at 0.025 Hz to evoke field excitatory postsynaptic potentials (fEPSPs) with a sample rate of 10,000 Hz. ripple frequencies of hippocampal CA1 region in awake rats (Suzuki and Smith, 1988;Ylinen et al., 1995). Moreover, such a firing pattern may play a role in physiological inhibitory regulation within the hippocampus and was shown to be more effective in triggering the ventral hippocampal LTD (Izaki et al., 2000).
Paired-pulse responses were examined by applying afferent stimulation in the form of two pulses of equal intensity and duration (0.2 ms) at interpulse intervals (IPIs) of 10, 20, 25, 50, and 100 ms. Individual pairs of stimuli were delivered at 40 s intervals, and individual IPI blocks of stimulation were delivered at 5 min intervals. Five stimulation pairs at each IPI were averaged and used for the analysis.
For the whole-cell current-clamp recordings, hippocampal sections from the dorsal, intermediate, and ventral subdivisions were continuously perfused in heated (32 8C) oxygenated aCSF (as described earlier). Our aCSF solution did not contain any agents that would influence the triggering fast EPSPs/IPSPs (e.g., picrotoxin). Pyramidal neurons in the middle of the proximodistal axis of the CA1 region were visualized at 403 magnification using an Olympus BX51WI microscope and an infrared video camera (TILL Photonics, Gräfelfing, Germany).
Recording patch electrodes (6-9 MX) were pulled from borosilicate glass with an external diameter of 1.5 mm using a Flaming/Brown micropipette puller (P-1000, Sutter Instruments, CA). Electrodes were filled with an intracellular solution containing (in mM: 97.5 K-gluconate; 32.5 KCl; 10 HEPES; 1 MgCl 2 ; 4 Na 2 ATP; 5 EGTA; pH 7.3). Whole-cell current-clamp recordings were performed on the soma of the CA1 pyramidal neurons, without a correction for liquid junction potentials, using a HEKA EPC10 amplifier and PATCHMASTER data acquisition software. Signals were low-pass filtered at 2.9 kHz and digitized at 10 kHz. Data were analyzed in an off-line mode in FITMASTER program.
The somatic input resistance (R in ) was measured as the slope of the voltage-current plot generated in a response to hyperpolarizing and depolarizing current injections (-80 to 1 20 pA, steps of 20 pA).
The membrane time constant was calculated as the slow component of an exponential fit of the averaged voltage decay in response to a hyperpolarizing current injection (-40 pA, 600 ms). Single action potentials were analyzed for action potential threshold (current and voltage), action potential amplitude, action potential half-width and afterhyperpolarization (AHP). Threshold (current) was defined as the current needed to induce an action potential. Threshold (voltage) was determined as the membrane voltage by reaching which an action potential is generated and was measured from the resting membrane potential (RMP). Action potential amplitude was measured as the voltage difference from threshold to peak, with the half-width measured at half this distance. AHP was determined as the voltage difference from the RMP to the peak of an undershoot. Firing frequency was calculated by averaging the instantaneous firing frequency of action potentials in a response to depolarizing current injections ranging from 50 to 400 pA.     (Table 1). However, they seem, at least in part, to depend on differences in threshold (voltage) between the ventral, dorsal, and intermediate neurons, whereby the ventral CA1 pyramidal cells exhibited significantly more depolarized values than the dorsal or intermediate ones  Table 1).

| D1 expression is the lowest in the dorsal CA1, while D2 is comparable across the dorsoventral axis
With regard to the other parameters measured, we also found a significant difference in the action potential amplitude between the ventral cells and neurons from the other two subdivisions (one-way ANOVA: F (2,46) 5 3.62, p 5 .03), but no change in the action potential half-width (one-way ANOVA: F (2,46) 5 0.9, p 5 .41), or in the membrane time constant (one-way ANOVA: F (2,46) 5 2.33, p 5 .1) ( Table 1).

| LTP and LTD profiles differ at the Schaffer collateral-CA1 synapses across the hippocampal longitudinal axis
To ascertain if differences could be identified on the level of synaptic plasticity between the dorsal, intermediate, and ventral parts of the CA1 region, two forms of long-term synaptic plasticity, namely LTP and LTD were examined. In order to elicit LTP, we applied TBS to hippocampal slices from induction was significantly higher than in ventral slices 10 min after the In summary, LTP was stronger in the dorsal pole and was equiva-

| Neurotransmitter release probability is the lowest in the dorsal hippocampus
To estimate if neurotransmitter release differs along the dorsoventral hippocampal axis, we then assessed neurotransmitter release probability at Schaffer collateral-CA1 synapses with the help of a widely used proxy method known as the paired-pulse response paradigm (Dobrunz & Stevens, 1997;Regehr, 2012). IPIs of 20, 25, 50, and 100 ms were used.
No differences were detected at the 100 ms IPI between any of the hippocampal parts (Figure 4d). These findings suggest that the intermediate-ventral two thirds of the hippocampal axis exhibit a higher neurotransmitter release probability compared with the dorsal third.

| D ISC USSION
This study describes a detailed characterization of the expression of plasticity-related receptors across the dorsal, intermediate and ventral subdivisions of the hippocampus. In addition we conducted a physiological comparison of neuronal properties and synaptic plasticity in these subdivisions.
Our key findings are the following: channel opening times that last approximately three times longer than that of GluN2A-containing receptors (Wyllie, Livesey, & Hardingham, 2013 to social and emotional information processing (Segal, Richter-Levin, & Maggio, 2010;Strange et al., 2014).
In contrast to all other receptors scrutinized here, that generally showed an expression gradient whereby the VH typically showed high-  (Bettler, Kaupmann, Mosbacher, & Gassmann, 2004) was shown to directly inhibit several types of voltage-sensitive calcium channels and generate inhibitory postsynaptic potentials via G protein-coupled inwardly rectifying K 1 (GIRK) channels (Chalifoux and Carter, 2011;Degro, Kulik, Booker, & Vida, 2015;Liu et al., 2012;Sun and Wu, 2009;Yang, Tadavarty, Xu, & Sastry, 2010). This suggests their greater contribution to a hyperpolarizing action of GABA A receptors, specifically in the VH. In turn, this would be expected to lead to a stronger inhibitory control and thus, lower excitability of the ventral CA1 principal cells.
In line with this interpretation, we showed that pyramidal neurons of the ventral CA1 responded with fewer action potentials, following current injections, than neurons in the dorsal and intermediate subdivi-  (Bean, 2007;Colbert & Pan, 2002). In agreement with this suggestion, a recent study showed that pyramidal neurons of the ventral CA1 region express significantly higher levels of Ca 21 -activated SK-type K 1 channels. These, in turn, were shown to inhibit NMDARdependent EPSP amplification to a greater degree in the ventral as opposed to the DH (Babiec, Jami, Guglietta, Chen, & O'Dell, 2017).
Neurotransmitter release properties were also altered along the dorsoventral axis. We saw stronger paired pulse facilitation in the dorsal CA1 following stimulation at intervals of 20, 25, and 50 ms compared with the ventral and intermediate parts. These findings align with reports by others for both the CA1 (Papatheodropoulos and Kostopou los, 2002) and CA3 regions of the hippocampus (Pofantis, Georgopoulos, Petrides, & Papatheodoropoulos, 2015) and suggest that the intermediate and ventral hippocampus display a higher release probability (Zucker and Regehr, 2002). GABAergic modulation influences paired-pulse facilitation by means of transient depression of postsynaptic inhibition (Davies, Davies, & Collingridge, 1990;Nathan and Lambert, 1991), albeit at IPIs of 100 ms or more. This would be expected to affect GABA B receptors (Gassmann and Bettler, 2012 (Maggio and Segal, 2007a,b;Papatheodoropoulos and Kostopoulos, 2000).
Mechanistically, such differences, in the magnitude of potentiation, may be related to the differential expression of plasticity-related receptors that we detected along the dorsoventral hippocampal axis as well as to reported differences in ion channel expression. In vivo, GluN2Acontaining NMDARs are important for the induction of the early phase of LTP (E-LTP, < 1 h), whereas activation of GluN1/GluN2B-containing NMDARs appear more important for LTP that lasts for longer periods (Ballesteros, Buschler, K€ ohr, & Manahan-Vaughan, 2016). GluN1/ GluN2B-containing receptors also require a higher membrane depolarization for their activation compared with GluN1/GluN2A-containing receptors (Clarke, Glasgoq, & Johnson, 2013 (Babiec et al., 2017), would be expected to escalate the propensity differences of hippocampal parts in expressing long-term synaptic potentiation. LTP maintenance is supported by mGlu5 (Mukherjee and Manahan-Vaughan, 2013) and this receptor also potentiates NMDAR currents (Doherty, Palmer, Henley, Collingridge, & Jane, 1997;Perroy et al., 2008). We found that mGlu5 was equivalently expressed across the hippocampal dorsoventral axis, however, suggesting that this receptor contributes little to the differences in LTP that we observed.
One interesting prediction that arises from the findings of this study is that the weaker LTP levels we observed seem to relate to a putatively stronger inhibitory control within the ventral pole of the hippocampus, which may not necessarily arise as a result of differences in GABAergic receptor expression, but rather may occur due to a more complex interaction between GABAergic, glutamatergic, and dopaminergic receptors, as well as ion channels. This suggests that information encoding by means of LTP does not readily occur in this structure. The higher levels of GluN1/GluN2B receptors suggest however, that under circumstances where this inhibitory control can be overcome, LTP will FIG URE 5 Overview of the hippocampal dorso-ventral CA1 region differences in receptor protein expression and synaptic plasticity levels. For all receptors, the "Baseline" level of expression refers to the lowest measured level of protein expression; a "Higher" level of expression refers to a significantly higher level of expression compared with the "Baseline" level; and the "Highest" level of expression corresponds to a significantly higher level of expression compared with the "Baseline" and "Higher" levels. not only be greater in magnitude but more robust in its persistency.
Assuming, in turn, that the VH processes the emotional context of memory (Segal et al., 2010;Strange et al., 2014) and that this is enabled by means of LTP (Whitlock, Heynen, Shuler, & Bear, 2006), this would suggest that strongly salient (emotional) experiences will result in quite robust and persistent encoding in this part of the hippocampus.

| C ONC LUSI ON S
In summary, this study demonstrates that GluN1, GluN2B, GABA A , and GABA B receptors, mGlu1, mGlu2/3, and dopamine D1 receptors are heterogeneously distributed across the dorsoventral CA1 region. Strikingly however, these differences take the form of a gradient whereby, GluN1, GluN2B, mGlu1, GABA B , and D1-receptors are expressed lowest in the dorsal CA1 and highest in the ventral CA1, whereas GABA A is expressed highest in the dorsal CA1 and lowest in the ventral CA1.
Differences in neuronal excitability, neurotransmitter release probability and synaptic plasticity appear, along with these expression gradients ( Figure 5). Taken together, these findings suggest that differences in the expression levels of plasticity-related receptors underlie functional distinctions in synaptic information storage along the dorsoventral hippocampal axis. These, in turn, may underlie the ascribed role of the different subdivisions of the dorsoventral hippocampal axis in information processing, learning and memory.

ACKNOWLEDGMENTS
We gratefully thank Olena Shchyglo, Ute Neubacher and Beate Krenzek for technical assistance and both Nadine Kollosch and Petra