Selective information routing by ventral hippocampal CA1 projection neurons

See allHide authors and affiliations

Science  01 May 2015:
Vol. 348, Issue 6234, pp. 560-563
DOI: 10.1126/science.aaa3245

How the brain sorts and routes messages

How do higher brain areas communicate with each other? Do they send out all computations equally to all target areas and leave the recipient to extract the needed and relevant information? Or does the transmitting region package and route computations differentially to distinct target areas, depending on the content? Ciocchi et al. found that the ventral hippocampus routes anxiety-related information preferentially to the prefrontal cortex and goal-related information preferentially to the nucleus accumbens. Hippocampal neurons with multiple projections were more involved in a variety of behavioral tasks and in memory consolidation.

Science, this issue p. 560


The hippocampus computes diverse information involving spatial memory, anxiety, or reward and directly projects to several brain areas. Are different computations transmitted to all downstream targets uniformly, or does the hippocampus selectively route information according to content and target region? By recording from ventral hippocampal CA1 neurons in rats during different behavioral tasks and determining axonal projections with optogenetics, we observed subsets of neurons changing firing at places of elevated anxiety or changing activity during goal approach. Anxiety-related firing was selectively increased in neurons projecting to the prefrontal cortex. Goal-directed firing was most prominent in neurons targeting the nucleus accumbens; and triple-projecting neurons, targeting the prefrontal cortex, amygdala, and nucleus accumbens, were most active during tasks and sharp wave/ripples. Thus, hippocampal neurons route distinct behavior-contingent information selectively to different target areas.

Learning processes require precise neuronal communication across distributed brain networks. The dorsal hippocampus is primarily associated with spatial navigation and episodic memory, whereas the ventral hippocampus is implicated in affective and motivated behaviors as well (1). Along the septotemporal axis of the hippocampus, CA1 projection neurons classically target the subiculum and entorhinal cortex (2). In the ventral CA1 hippocampus (vCA1), additional projections have been described (2), such as to the medial prefrontal cortex (mPFC), nucleus accumbens (Acb), and amygdala (Amy). We asked whether the hippocampus transmits all of its computations equally to those target areas, ceding their interpretation to those areas, or rather routes distinct representations selectively to different brain areas according to information content?

We recorded the activity of vCA1 neurons from rats (n = 4) during an anxiety task, spatial exploration, and goal-directed navigation, all within single experimental sessions (Fig. 1A). After each session, we determined the projections of recorded cells with optogenetics (3). For this, vCA1 was previously (4 to 5 weeks before) injected with an adeno-associated virus expressing channelrhodopsin-2 fused to enhanced yellow fluorescent protein (ChR2-eYFP) under the control of a neuron-specific promoter. In addition to vCA1 somata and dendrites, ChR2-eYFP was expressed in vCA1 projection fibers and synaptic boutons targeting the mPFC, Acb, or Amy (Fig. 1B). Blue light pulses were delivered via optic fibers placed above the mPFC, Acb, or Amy, while recording single-unit activity and antidromically evoking spikes in vCA1 (Fig. 1A and fig. S1). Antidromic spikes were defined by early-latency, low-spike jitter and a high-fidelity response to light (Fig. 1, C and F, and figs. S2 to S4). Spike-collision tests (Fig. 1C, right) and entrainment by 50-Hz photostimulation (Fig. 1D) were also used. In control experiments under anesthesia, local application of d-(-)-2-amino-5-phosphonopentanoic acid (D-AP5, a competitive N-methyl-d-aspartate receptor antagonist) and 2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzo[f]quinoxaline-7-sulfonamide (NBQX, a competitive AMPA receptor antagonist) preserved antidromic but not spontaneous spiking (Fig. 1E and fig. S5), indicating that antidromic spikes are not generated via multisynaptic pathways. The spike waveforms of antidromic and spontaneous spikes recorded during behavior were similar (fig. S6). Using stringent criteria (fig. S7), we identified 99 out of 233 units (42%) as vCA1 projection neurons targeting the Amy, Acb, and/or mPFC.

Fig. 1 Optogenetic identification of vCA1 projection neurons.

(A) Behavioral tasks and antidromic activation of single units in vCA1 by photostimulating ChR2-expressing vCA1 axons in the mPFC, Acb, or Amy. (B) ChR2-immunopositive vCA1 axons targeting the mPFC, Acb, and Amy. Scale bar, 500 μm; in insets, 10 μm. Cg, cingulate cortex; PrL, prelimbic cortex; LS, lateral septum; Acbsh, nucleus accumbens shell; Ce, central amygdala; BLA, basolateral amygdala; BM, basomedial amygdala. (C) Low-jitter, high-fidelity, and early-latency response of a vCA1 neuron upon light stimulation (blue) in the Acb. Spike collision is shown at trial 18. (D) Reliable firing after 50-Hz stimulation. (E) Local infusion of NBQX/D-AP5 (APV) reduces spontaneous but not light-induced firing. (F) Spike latency, jitter, and fidelity in vCA1 projections targeting the Amy (n = 21 neurons), Acb (n = 79), or mPFC (n = 52). Plots indicate median, quartiles, and range.

During recording sessions, naïve rats were placed onto an elevated plus maze with two open and two closed arms. Rats spent less time in and made fewer entries into open arms (more anxiogenic) versus closed arms (less anxiogenic, fig. S8A). We discovered that a subset of vCA1 neurons fired with different frequency whenever the rat was located in open arms as compared to closed arms (Fig. 2, A and B). Their firing rates were positively correlated in open and closed arms but negatively correlated across different types of arm (fig. S8, B and C). When the positions of closed and open arms were exchanged, the neurons maintained their firing respective to anxiety levels (Fig. 2, A and B, and fig. S8D). Using an elevated plus maze/neurophysiological score for such firing patterns (4), we found more neurons with anxiety-related firing in vCA1 than in the dorsal CA1 hippocampus (dCA1, Fig. 2C). Using a bootstrap analysis (see supplementary methods and results) and combining for each neuron elevated plus maze scores and axonal projection sites, we observed that the vCA1-to-mPFC projection was selectively enriched with vCA1 anxiety-related neurons (Fig. 2C). Absolute levels of anxiety-related firing were larger in vCA1 neurons projecting to the mPFC than in those targeting the Amy (shared projections to both targets not counted, Fig. 2D). Most cells with anxiety-related firing had higher firing in the open arm (fig. S8E), and elevated plus maze scores of individual neurons correlated with the frequency of visited open/closed arms (fig. S8F).

Fig. 2 vCA1 neurons projecting to the mPFC are enriched with information about anxiety.

(A and B) Firing of individual vCA1 neurons on an elevated plus maze with changes of maze configurations. (C) Percentage of anxiety cells in dCA1, vCA1, and identified vCA1 projection neurons. vCA1 → mPFC projections exhibit a larger number of anxiety cells as compared to chance. Numbers depict the ratio of anxiety cells. (D) Elevated plus maze (EPM) scores (mean ± SEM) of vCA1 neurons projecting to the mPFC or Amy (without axon collaterals to both). (E) Fraction of anxiety, place, and overlapping anxiety/place cells. (F) Firing-rate maps of dCA1 and vCA1 place cells. Scale bar, 50 cm. (G) Percentage of place cells in dCA1, vCA1, and identified vCA1 projection neurons. *P < 0.05, ***P < 0.001.

Next, rats were placed in a large open field (180 × 180 cm) to detect place cells in vCA1. Consistent with the reported lower spatial information content across the septotemporal axis of the hippocampus (5, 6), vCA1 place cells had larger place fields (Fig. 2F and fig. S9) and less spatial tuning and stability than dCA1 place cells (fig. S9). Place cells in vCA1 were less numerous than dCA1 place cells (Fig. 2G); vCA1 anxiety-related neurons had little overlap with place cells (Fig. 2E). We found that vCA1 projections carrying spatial information directly targeted the Amy, Acb, and mPFC, although none of the projection types were significantly enriched with place cells (Fig. 2G).

Next, rats performed a goal-directed navigation with changing rules (fig. S10A). None of the vCA1 projection types were specifically recruited after a rule change. However, we observed that a subset of vCA1 neurons increased firing, whereas another subset decreased firing when animals approached the reward zone (Fig. 3, A and B). Changes in firing rates were progressive, observed across behavioral conditions (fig. S10B), and occurred before the entry into the reward zone, suggesting that these firing patterns were not simply a mere consequence of reward consumption. Firing rates of vCA1 neurons were not correlated with speed or acceleration (fig. S11). Reward zone–excited neurons were preferentially observed among vCA1 double projections targeting the mPFC and Acb, whereas reward zone–inhibited neurons were enriched in vCA1 projections targeting the Acb (Fig. 3, C and D). The firing of vCA1 neurons projecting to the Acb was more strongly modulated during entry into reward zone compared to other projection neurons (Fig. 3E and fig. S12).

Fig. 3 Goal-directed firing is preferentially carried by vCA1 → Acb projections.

(A and B) (Top) Firing of a reward zone–excited neuron and a reward zone–inhibited neuron during a goal-directed navigation task. (Bottom) Significant changes in firing precede entry into a reward zone by 1.1 or 0.9 s in population histograms (mean ± SEM). (C) (Top) Percentage of reward zone–excited and reward zone–inhibited neurons among vCA1 projection types. Neurons projecting to the mPFC and Acb are preferentially enriched in reward zone–excited neurons, whereas vCA1 → Acb projections are preferentially enriched in reward zone–inhibited neurons. (D) Proportions of reward zone–excited and reward zone–inhibited neurons among vCA1 → Acb/mPFC and vCA1 → Acb projections. (E) Absolute difference (mean ± SEM) of firing in reward zone and maze for neurons with or without projections to the Acb. *P < 0.05, n.s.: not significant.

We observed that neurons with triple projections targeting the mPFC, Amy, and Acb showed the most prominent behavior-dependent firing during the three tasks altogether and exhibited higher firing during behavioral tasks (Fig. 4A and fig. S13). The higher firing of these cells was also present before task performance (Fig. 4B). We observed that vCA1 triple projections were recruited most during sharp wave/ripples (Fig. 4C), suggesting that vCA1 triple projections contribute chiefly to vCA1-dependent memory consolidation (7).

Fig. 4 Preferential recruitment of vCA1 triple projections during behavior and sharp wave/ripples.

(A) (Left) Higher overall behavioral responsiveness of vCA1 triple projections as compared to chance and (right) to other pooled vCA1 projections. Numbers indicate the ratio of task responses in each group (a neuron responding to three tasks counts 3; for alternative plotting, see fig. S13). (B) (Left) Higher firing rates (mean ± SEM) of vCA1 triple projection cells during behavioral tasks (non-identified projections include interneurons). (Right) Pre-task and during-task firing are positively correlated across projection types. (C) (Left) Preferential recruitment of triple projections during sharp wave/ripples. SWR, sharp wave/ripples. Numbers indicate the ratio of sharp wave/ripples to active neurons in each group. (Right) Firing of neurons during sharp wave/ripples. Scale bar, 40 ms, 100 μV. (D) Schematic showing preferential information content of identified vCA1 projection neurons. *P < 0.05.

We have shown that specialized projection pathways in vCA1 route behaviorally relevant information to distinct neural networks (Fig. 4D, fig. S14, and table S1). Our application of strict conditions to accept antidromic spike activation and the lack of synaptic inputs to vCA1 from the Acb and mPFC suggest a reliable classification of projection types. Although ChR2 was universally expressed in vCA1 neurons at the recording sites, we cannot exclude the possibility that some projections remained undetected with our method. However, such unaccounted projections could not generate the significant differences between cell groups described here, and any unaccounted-for projections within our data set would have led to an underestimation of observed differences. Also, despite the behavioral control experiments performed, it can never be fully excluded that the anxiety- and goal-related firing described here may reflect more complex aspects and computations of the vCA1 network.

The mPFC and Amy are involved in anxiety behavior, receiving direct inputs from vCA1 (2, 4, 8). We demonstrated that anxiety-related activity is preferentially supported by vCA1 → mPFC projections, in agreement with described theta-frequency synchronization between the ventral hippocampus and mPFC during anxiety behavior (9). Nevertheless, the Amy could receive anxiety-related signals indirectly and processed them via the mPFC (10, 11). We hypothesize that vCA1 → Amy projections may rather contribute to contextual fear memories (12). Our results support a differential contribution of the dorsal and ventral hippocampus to spatial and anxiety behaviors (1, 1315). Neural representations of space and anxiety coexist in vCA1 but are conveyed by distinct vCA1 projection types, which may receive segregated space and anxiety inputs from the Amy (16) or entorhinal cortex (17, 18). Alternatively, this segregation could be boosted by local parvalbumin-positive basket cells, which differentially inhibit CA1 projections targeting the Amy or mPFC (19). Additionally, projection type–specific plasticity could fine-tune the formation of place or anxiety neurons in vCA1 (2023). Context-dependent fear renewal, conditioned place preference, or spatial working memory require spatial information to reach the Amy, Acb, or mPFC, respectively (2426). We have demonstrated that place cells among vCA1 projection neurons indiscriminately target these areas and may support spatially driven cognitive processes. The wide-ranging presence of spatial information along the septotemporal axis of the hippocampus may coordinate the expression of interference and generalization, pertaining to mnemonic processes (27, 28).

We found two types of neuronal response among vCA1 projection neurons, with consistent trial-by-trial discharges in anticipation of reward outcomes, which were observed under numerous behavioral conditions, suggesting that this may be a universal phenomenon among subsets of vCA1 projection neurons. Goal-directed firing is conveyed to the Acb and mPFC by distinct vCA1 projections and may tune corticostriatal loops for goal-directed behavior (29, 30).

Our results indicate that higher cortical areas, such as the vCA1, communicate with other brain areas not by transmitting all of their computations equally but by routing the information according to content and recipient.

Supplementary Materials

Materials and Methods

Figs. S1 to S14

Table S1

References (3138)


  1. Acknowledgments: We thank P. Schönenberger [Institute of Science and Technology (IST), Klosterneuburg, Austria] and S. Wolff (Harvard Medical School, Boston, USA) for helpful discussions on the optogenetic strategy; T. Asenov (IST) for three-dimensional printing of microdrives; R. Tomioka (Kumamoto University, Japan) for help in setting up the behavioral and electrophysiological experiments; E. Borok and R. Hauer for technical help with histology; all the members of the Klausberger lab for insightful discussions; and P. Somogyi, T. Viney, and M. Lagler for commenting on an earlier version of the manuscript. We thank Penn Vector Core from the University of Pennsylvania for the adeno-associated virus (AAV). The use of the AAV is disclosed by an materials transfer agreement between the University of Pennsylvania and the Medical University of Vienna. This work was supported in part by grant 242689 of the European Research Council and grant SCIC03 of the Vienna Science and Technology Fund. The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. Some of the original data are shown in the supplementary materials; all other data are available upon request from the corresponding authors.
View Abstract

Navigate This Article