Two opposing hippocampus to prefrontal cortex pathways for the control of approach and avoidance behaviour

The decision to either approach or avoid a potentially threatening environment is thought to rely upon the coordinated activity of heterogeneous neural populations in the hippocampus and prefrontal cortex (PFC). However, how this circuitry is organized to flexibly promote both approach or avoidance at different times has remained elusive. Here, we show that the hippocampal projection to PFC is composed of two parallel circuits located in the superficial or deep pyramidal layers of the CA1/subiculum border. These circuits have unique upstream and downstream connectivity, and are differentially active during approach and avoidance behaviour. The superficial population is preferentially connected to widespread PFC inhibitory interneurons, and its activation promotes exploration; while the deep circuit is connected to PFC pyramidal neurons and fast spiking interneurons, and its activation promotes avoidance. Together this provides a mechanism for regulation of behaviour during approach avoidance conflict: through two specialized, parallel circuits that allow bidirectional hippocampal control of PFC.

Major 1-The exact nature of PFC-projecting cells is unclear. The region from where these cells are examined corresponds to the distal aspect of the ventral CA1 at the border with the subiculum. According to the Hippocampus Gene Expression Atlas (Bienkowski et al., 2018) and cell-type specific gene mapping (elsewhere) these cells are not necessarily CA1 but a mix of distal CA1 and subicular cells. The location and distribution shown in Fig.1b and their bursting/non-bursting phenotype (Fig.1e,f) are more consistent with subicular cells or at least of cells recorded at the border. Actually, the segregated distribution of deep and superficial cells in Fig.1b,i reminds those subicular layers 2 and 4 (according to HGEA), and not the typical deep/superficial continuum of packed CA1. Making horizontal slices complicates interpretation due to the different spatial orientation of the different hippocampal axes and alignment with validated regions is not so straightforward. Thus, it would be very helpful to validate the CA1 or subicular nature of PFCprojecting cells by using immunostaining (IH) and/or any other complementary assay. While I don't feel this as a limiting factor, the whole paper would benefit from being more rigorous in the real nature of these cells.
2-Authors use Calb-cre mice to target superficial CA1 cells. While this is the case for the dorsal hippocampus, it is unclear how calb expression relates to superficial CA1 cells at more ventral regions. It is also unclear how calb is expressed in the subiculum. In addition, authors should provide validation of their Calb-Cre on/off strategy. Finally, calbindin is also expressed in interneurons and in dentate granule cells and authors need to exclude any potential contribution/contamination. For interneurons in particular, calbinding expressing interneurons may be projecting cells. This issue is relevant as many of the viral strategies uses ubiquitous promoters. Thus, overall, the paper lacks detailed anatomical validation and support.
3-Authors use fiber photometry to evaluate activity of PFC-projecting deep and superficial cells during EPM exploration. While most analysis concentrate in evaluating transient events, the basal level of optical signal is not discussed nor shown. Only 4 sec of entry on the arm are analyzed so, it is unclear what is the typical activity along the entry and how does it evolve over exploration. Raw data and validation of a stable basal level should be provided as supplementary information. It is also unclear whether the fiber is sensing bulk activity from dispersed soma (as suggested in previous figures) and how precisely can deep and superficial cells be separated given their different density (and hence contribution to bulk imaging). Are fibers of passage or dendrites of deeper cells contaminating superficial signals? In addition, how can authors precisely define the entry of mice into the arm? All these experiments require substantial support from the anatomical and analytic point of views.
4-Authors use optogenetics to manipulate the activity of deep and superficial PFC-projecting cells while mice explores the EPM. I am not fully convinced the behavioral effects are clean and statistically strong. First, it is unclear how the protocol goes. Authors light on only during the 3 min period in the central area or in the open arms. what if an animal stays in the closed arm? What proportion of animals receive lights actually? Second, there is lot of variability both in Cre-ON and GFP-control animals regarding the time in open arm, so it is unclear what the normalized time is doing here to reach significant statistical effects. All animals exhibit a decrease of exploration to the open arm (and hence they should be increasing exploration of the closed arm) along the task except for the CreON group, and some individual GFP-control mice. Thus, the effect of optogenetic looks a bit at the limit. Authors need to show convincingly that the behavioral effect is robust. Is stim in the closed arm exerting a complimentary effect that could reinforce statistics and conclusions?
Other comments 5-Regarding quantification in Fig.1: A) Please clarify whether quantification of radial distance in Fig.1b refers to several sections or is just one representative section; B) Please, clarify whether cells are counted in a stack (70 um thick) or at a plane; how many planes? What objective was used to quantify? C) In methods authors refer to "either the transverse, coronal or sagittal planes as described in the figure legends", but most examples look horizontal or sagittal. No coronal sections are presented which would facilitate clarification of my point 1 (HGEA are represented in coronal sections). D) Fig.1c and Supp.Fig.1d: authors refer to pia, but strictly speaking, this is the alveus. The pia is the layer of meninges surrounding the surface of the brain, which is more likely to be at the hippocampal fissure. E) Fig.1i, please improve contrast of the fluorescent image 6-Check arrowheads in sup. Fig.2b; some point erroneously to distinct cells. Fig.2f: the retrobead injection is lacking. According to the text retrobead-negative cells are recorded here 8-Authors declare "Excitatory local connectivity in hippocampal CA1 and subiculum is rare" but this is not strictly correct. Not certainly at the ventral subiculum, nor between CA1 and subicular cells. Fig.2g,h,I exploits somatic ChR2 but no supporting material is providing to validate their resolution. In addition, can authors fully exclude the fibers of passage? Also validation of interneuron-specific expression of mRuby (and electrophysiological validation) should be provided. A supplementary figure is critical here 10-For the quantification of PFC-projecting cells in Fig.3a,b: it is unclear why authors present data as proportion of cells in the superficial layers and not along the normalized distance as they reported before. Interpreting this plot is very tricky. Also, in the table author refer the n for deep and sup which unclearly relate to CamKII vs VGAT (as shown in Fig.3b). Fig.3c, authors need to show the distribution of the axonal plexus in the PFC to clarify whether deep and sup cells are actually innervating different layers and/or dendritic domains which could additionally complicate interpretation. Interestingly, authors record only from deep PFC cells, while Lee et al., Neuron 2019 (ref 44) record from layers II/III. Fig.3f: what PFC layers are recorded here? All this should be more carefully considered and justified. 12-Supp. Fig.3: the injection site and fiber placement: please, show some examples to appreciate the size and potential anatomical effect of the fiber implantation.

11-Regarding axonal arborization of PFC-projecting cells in
13-I strongly recommend separating fiber photometry and optogenetic behavioral experiments in different figures, and adding substantially more analysis, and support to results.
14-Depending on their ability to address my point 1 I would recommend authors making more clear the CA1/subicular nature of the PCF-projecting cells and shape title and abstract accordingly. 15-Discussion, in general reads a bit oversimplified. I would have several comments and additions but in view of the overall assessment and the many details authors need to work out to improve this interesting paper I would save my comments for a revised version.
Reviewer #2 (Remarks to the Author): In their manuscript, Sánchez-Bellot and MacAskill examine the local-and long-range connectivity of two circuits connecting the ventral hippocampus to the prefrontal cortex, which differ based on radial depth of ventral hippocampus neurons, and relate these circuits to approach/avoidance behavior. They associate activity in the deep ventral hippocampus circuit with avoidance behavior and activity in the superficial ventral hippocampus circuit with exploration/approach. Finally, they show that optogenetic activation of each of these circuits leads to the predicted behavior, demonstrating a causal mechanism by which the hippocampus plays a bidirectional role in approach/avoidance behavior.
The experiments presented in the manuscript are challenging and well-executed. The paper does an outstanding job of bridging current coarse scientific knowledge into fine-scale discovery: the authors hinge their experiments on previously identified cell types (i.e., superficial vs. deep ventral hippocampus neurons) and a known bidirectional role for the hippocampus in approach/avoidance, but are the first to show that this hippocampus function derives from distinct circuits involving these cell types. As such, the paper fills a clear knowledge gap and is timely. Given their novel findings spanning multiple scales (i.e., circuit mapping to behavior), recapitulation and extension of previous results, and the potential eventual translational relevance of their results, this manuscript is of interest to broad audience and well-suited to a strong journal like Nature Communications.
In this reviewer's opinion, no major revisions nor further experiments are needed. However, the clarity and convincingness of some parts of the paper could be improved by considering the following minor comments and addressing via additional discussion or rewording.
Minor comments: • CRACM-based conclusions: I would suggest that the authors exhibit more restraint in discussing their CRACM-associated conclusions for Fig. 2. Their recordings are not well-positioned to compare inputs across superficial vs. deep cells for at least two reasons. First, morphological differences will mean that inputs -especially distal inputs -will have differential somatic signatures. Second, recordings were performed in current clamp, which means that cell-intrinsic differences may distort synaptic inputs. Thus, more equivocal wording should be employed -claims like "additional input" can be suggested by not truly resolved by these experiments. • Off-target effects, optogenetic stimulation: Simulation of axon terminals can result in antidromic spiking back to the cell body, which can then orthodromically propagated to other axon terminals that are associated with different collateralizeed circuits. This off-target stimulation of collateralized circuitry can be a potential confound. The authors should discuss this possibility in the context of their experimental results. • CA1 vs subiculum: While the authors state that the neurons they study are at the "CA1/subiculum" border, they do not attempt to describe which brain region (CA1 vs subiculum) these neurons belong to. Given that the exact boundary between the CA1 and subiculum has been debated in the field (especially in ventral hippocampus, as in this study), this is a challenging task. Nonetheless, some discussion as to what hippocampal region these cells reflect, or indeed if this is even possible to interpret the context of the authors' experiments, is important. As many in the community think about these outputs in terms of CA1 vs. subiculum, it will be important for this paper to discuss their results in the context of these regions. • CreON vs. creOFF explanation: The wording describing the creON/creOFF experiment was confusing and seemed contradictory: " [We] injected an AAV into vH to express either ChR2 where expression was limited to cre-expressing (superficial layer) neurons (creON ChR2), or where ChR2 was inhibited in cre-expressing neurons, and thus was only expressed in cre-negative (deep layer) neurons," as the sentence structure makes it seem as though in both creON and creOFF ChR2 is expressed in only cre-negative neurons.  Figure 3c is (it is provided in 3f, but would make more sense to have this legend in the first panel the reader views). In the current text, is also unclear how this schematized neuron is related to the experiment. • Reporting of cell sample sizes: Often, the number of superficial vs. deep ventral hippocampus and PFC analyzed cells are not reported in the text or on relevant figures (e.g., Figure 1gpercentage of bursting cells is reported for both superficial and deep hippocampus cells, but not the actual number of cells examined; although there are dots for individual cells, the burden is on the reader to manually add the data points). Please report sample sizes throughout, as possible.
Reviewer #3 (Remarks to the Author): Sanchez-Bellot & MacAskill used a combination of in vitro slice electrophysiology, anatomical tracing, in vivo fibre photometry and optogenetics to identify two populations of medial prefrontal cortex (mPFC)-projecting ventral hippocampal (vH) cells that have distinct electrophysiological and projection profiles, as well as opposing control over the expression of innate approach-avoidance conflict, as measured in the elevated plus maze (EPM). The major findings of the study are that vH cells in the superficial layers in the subiculum and CA1 receive medial entorhinal inputs, and connect to both somatic and dendritic inhibitory neurons in the PFC in order to facilitate approach/exploratory behavior. In contrast, vH cells residing in the deep layers receive subcortical projections (anterior thalamus, DBB) and project onto somatic interneurons and pyramidal cells in the PFC to promote avoidance behavior. This is a very interesting set of experiments that have been conducted carefully and thoroughly. The findings provide novel mechanistic insights into the control of anxiety, and will likely be of interest to an audience interested in the neurobiology of anxiety, and function of hippocampal and prefrontal circuits. The statistical analyses are appropriate, and the methodology is clearly written and reproducible on the whole, although more clarification and details are warranted in some parts, as described below. I have also outlined some concerns that would need to be addressed in order for me to fully assess the validity of the claims.
I appreciate the reason why the authors chose to stimulate contralateral vCA3 inputs to the PFC projecting cells, but this raises the question of whether there may be intrinsic differences in the projection pattern/physiological profile of the neuronal population targeted by ipsilateral vs. contralateral vCA3 inputs, and whether the omission of recording from the ipsilateral pathway in the current study is an issue.
Did the authors assess the collateral projections of the superficial and deep layer PFC-projecting vH cells?
The behavioral and data analyses that the authors conducted are not best suited to justifying some of the claims the authors are making. For instance, the authors claim their fiber photometry findings "suggested that the relative activity of the superficial and deep layers of the vH-PFC projection around the choice point of the EPM may inform the decision to approach or avoid the open arms." However, neural activity was analysed during 0 to 4 seconds after arm entry, and changes in activity also appear most prominent after the entry into either the open or closed arm (Figure 4 c,d). Given that approach and avoidance decisions likely occur before the animal enters the arm, and not after, it is possible that the time window the authors include in their analysis may not be representative of decision-making. Was neural activity analysed for the 2s preceding arm entry (-2s to 0s)? series data was then calculated for each behavioral session as the z-score of (green or red signalauthors also state, "Activity was analyzed as the area under the curve from 0 to 4 s after arm entry, normalized to the area under the curve -4 to -2 s before arm entry". I assume this resulted Please provide details as to how many open and closed arm entries were included in the data analysis. Did the observed changes in activity differ between successive entries into the open or closed arms (those made at the beginning vs. end of session)?
It is of concern that the placements of optic fibres in the fibre photometry experiments seem to be much more posterior and dorsal to the sites targeted in the tracing and in vitro slice experiments (Supp Fig 3b).
The inclusion of the optogenetic activation experiment is a nice way of demonstrating the sufficiency of the differential vH-mPFC circuits in approach/avoidance behaviors. However, it is not a demonstration of the necessity of these circuits in the control of anxiety -did the authors examine the effects of inhibiting these circuits on approach/avoidance? Perhaps the manuscript is intended to be a short report, but I found the discussion to be lacking, and have a few suggestions for improvements. Firstly, the authors should discuss their findings in the light of recent work that explore the contributions of different vH subfields in innate and cued approach-avoidance behaviors (e.g., PMID: 25931556, PMID: 29606418, PMID: 32929245). Secondly, it's not clear how exactly the reported circuit mechanisms (particularly the contrasting innervation patterns of the mPFC) give rise to the implementation of preferential approach vs. avoidance in a conflict situation. Furthermore, huge conceptual leaps are made in stating that superficial neurons use structured information from the cortex to plan and promote approach/exploratory behavior, and deep layer neurons use salient information in the environment flexibly to promote avoidance. The authors should better articulate/elaborate on how they have arrived at these conclusions.
Please indicate if, and which within subject post-hoc comparisons were significant in the EPM data ( Figure 4