Medial entorhinal VIP-expressing interneurons receive input from the thalamic anterior dorsal nucleus and are critical for spatial memory

Significance

Interneurons are classically viewed as essential components in the maintenance and regulation of the excitatory/inhibitory balance in local neuronal networks. Nevertheless, recent studies have shed new light on vasoactive intestinal peptide cells (VIP), a subtype of interneurons, as a key recipient of long-range projections underlying their active role in propagating and processing information in brain networks. In this report, we highlight the rich connectivity of medial entorhinal cortex VIP cells, as these cells receive strong inputs from space- and memory-related brain regions, including the anterodorsal thalamic nucleus (ADN). In this line, we also show evidence of the critical role of medial cortical VIP cells in memory.

Abstract

Vasoactive intestinal peptide (VIP) cells have emerged as a crucial component of the inhibitory network, facilitating interregional communication, particularly in the context of associative memory. To expand our current understanding of the role of this population, we combine anatomical and functional approaches to question the influence of VIP cells in the medial entorhinal cortex (MEC), a region key for navigation and spatial memory. From our anatomical tracing study, our findings reveal that neurons located in the anterodorsal thalamic nucleus (ADN) specifically project onto MEC interneurons, with a strong preference for MEC VIP cells. Additionally, MEC VIP cells receive input from the hippocampus, the subicular complex, and the retrosplenial cortex, suggesting a specialized role for MEC VIP cells in spatial memory. Indeed, we find that MEC VIP cells exhibit increased c-Fos expression in a spatial memory task and show that chemogenetic inhibition of these neurons impairs task performance. Together, these data reveal a specific projection of head-direction (HD) information onto MEC interneurons and confirm that MEC VIP-expressing cells play a critical role in spatial memory.

The medial entorhinal cortex (MEC) is essential for spatial navigation and memory (1–3). Within the MEC microcircuit, excitatory cells in the superficial LII-LIII MEC layers exhibit various profiles of spatial firing (3), such as grid cells (4), border cells (5, 6), head-direction (HD) cells (7–9), or object-vector cells (10), whose coordinated activity supports the neural representation of space. While tremendous advances have been made in understanding excitatory cells at the physiological (11–14), functional (11–15), and behavioral (11–13) levels, less is known about how interneurons regulate the information flow necessary for the emergence of space coding in the MEC (11–13, 16–19). The MEC is rich in inhibitory cells, such as parvalbumin (PV) (17, 20), somatostatin (SOM) (20, 21), or vasoactive intestinal peptide (VIP) cells (22), embedding the excitatory cells (17, 20, 23) in a network of inhibition. In contrast to PV and SOM cells, which directly project onto excitatory cells (1, 19), VIP cells are mostly disinhibitory (24–28) in most cortical structures. These interneurons either project onto PV or SOM cells and thus indirectly orchestrate the activity of excitatory cells by releasing them from ongoing inhibition. In the MEC, pharmacogenetic inactivation of PV cells resulted in impairment in the periodicity of grid cells (17). In contrast, the inactivation of SOM cells altered the spatial selectivity of aperiodic cells (17) but left border or HD activity unaffected (17). While MEC VIP cells’ firing and morphological characteristics have been reported (22), less is known about their role in entorhinal spatial coding properties. Despite limited knowledge, recent work highlights VIP cells as critical for memory in the auditory cortex (26), the amygdala (24, 26), and the hippocampus (29, 30). Interestingly, recent evidence demonstrates that hippocampal VIP cells are functionally versatile as they not only modulate principal cells’ firing but also adapt their activity according to task demands (29), supporting the hypothesis that interneurons integrate external information for spatially guided behavior. Moreover, VIP cells in the prefrontal cortex (31), the anterior insular cortex (32), and the amygdala (24) relay long-range inputs in their local network and control information flow to regulate behavior supported by these brain regions. Therefore, among the diverse interneuron subpopulations, VIP cells display a unique coordinator function that is central to shaping behaviors. These observations prompt the extension of the mapping of the anatomical projections to MEC VIP cells. Such cartography is essential for understanding the functional role of MEC VIP cells in relation to the variety of space-coding excitatory cells in the MEC. Hence, we hypothesized that MEC VIP cells, as observed in other brain regions, would receive external inputs from brain regions involved in spatial processing and would therefore be necessary for cognitive functions supported by the MEC, such as spatial memory.

Results

VIP Cells Receive Inputs From Key Memory and Space Brain Centers.

VIP^cre mice were first unilaterally injected with the helper virus AAV2/8-hSyn-FLEX-TVA-P2A-eGFP-2A-oG (TVA-eGFP) in the MEC and four weeks later received the RABV-DeltaG-EnvA-mCherry rabies virus (Fig. 1A). This procedure allowed for the localization and quantification of input cells (mCherry+), which made synaptic connections with the starter cells (GFP+/mCherry+) (Fig. 1 B and C). For quantification analysis, we included brain regions with at least one input cell in both planes (identified by the presence of mCherry+ cell bodies). The relative quantification of inputs received by individual starter cells (called the convergence index (CI)) was calculated by dividing the number of MEC starter cells by the number of input cells in a given brain region. In addition to this, we also quantified the fraction of inputs a brain region sends to starter cells in the MEC (total number of inputs for a ROI divided by the total number of presynaptic inputs in the entire brain) for each animal. VIP starter cells were observed across all the layers of the MEC, with a majority of them localized in the superficial layers and fewer in the deep layers (Fig. 1C and SI Appendix, Table S1, superficial: 48.75 ± 10.94, deep: 17 ± 5.67). This observation is consistent with a previous report identifying more VIP cells in the superficial layers (22). Quantitative analysis showed that VIP cells are highly connected within MEC microcircuitry, as most of the input cells were located within the structure (Fig. 1D and SI Appendix, Table S1, CI = 17.33 ± 6.57), especially the superficial layers (superficial CI = 11.98 ± 4.48, deep CI = 5.35 ± 2.09). These cells also received large projections from CA1 (Fig. 1 E–H and SI Appendix, Table S1, CI = 8.08 ± 4.46) and the subicular complex (Fig. 1 E–H and SI Appendix, Table S1, CI = 5.22 ± 1.50). In addition, we assessed the efficiency and specificity of TVA viral transfection (Fig. 1 I–K). Injection of the TVA-eGFP virus into the MEC of VIP-TdTom mice is shown in Fig. 1I. The specificity of the TVA virus was also confirmed by the absence of TVA-eGFP expression when injected into cre-negative animals (Fig. 1J). To quantify transfection efficiency, we further analyzed TVA-eGFP expression in VIP-TdTom mice (Fig. 1K), which revealed that approximately 60.9% of tdTom+ cells expressed GFP, while 97.5% of GFP⁺ cells also expressed tdTom (GFP: 35.13 ± 8.62; tdTom: 57.63 ± 3.87; GFP+/tdTom+: 34.25 ± 8.25). These results confirm that TVA-eGFP expression was primarily restricted to VIP interneurons in a cre-dependent manner. Among the cortical regions, dense afferents originating from the perirhinal (Fig. 1 E and L, CI = 0.028 ± 0.01) and the retrosplenial (Fig. 1 F, H, and L, CI = 1.10 ± 0.51, 3.05% of the total input cells) cortices were found. In contrast, sparse afferents came from the lateral entorhinal cortex (LEC) (SI Appendix, Fig. S3). Overall, the fraction of inputs from cortical areas known to process spatial memory constituted the largest population of projecting neurons (Fig. 1H, MEC: 52.05%, CA1 21.71%, and subicular complex 16.41%) as 90.17% of labeled soma were in one of these regions. In the forebrain, input cells were also identified in the medial septum/horizontal diagonal band (Fig. 1 L and M and SI Appendix, Table S1, CI = 0.38 ± 0.05). While sporadic input cells were found in the thalamus, most of the input cells’ soma were detected in the ADN (Figs. 1L and 2E, CI = 1.68 ± 0.68). Nevertheless, ADN contributes to 5.05% of the total projections received by MEC VIP cells (Fig. 1H). Residual input cells were also observed from the amygdala or the visual and auditory cortices (Fig. 1L). This result demonstrates that MEC VIP cells are highly interconnected within the MEC microcircuit. Meanwhile, this cell population also receives a significantly broader spectrum of inputs, including major HD coding centers such as the ADN and the subicular complex, but also regions essential to memory, such as the hippocampus, the retrosplenial, and the perirhinal cortices.

Fig. 1.

Tracing local and long-range projections to VIP cells in the MEC. (A) Strategy for helper virus and rabies virus injection to enable retrograde tracing of inputs to the MEC. The helper virus (TVA-eGFP) was injected into a VIP^cre mouse first, and then four weeks later, the rabies virus (RV-mCherry) was injected. The animals were perfused a week later for whole brain sectioning. (B) Image of the injection site and starter cells in VIP^cre mouse as seen in the (i) horizontal plane and (ii) sagittal plane. Dashed lines mark layers of the MEC. Yellow boxes and white arrows denote a magnified view of the starter and input cells within the MEC. Transfection of VIP cell by (iii) TVA receptors (TVA-eGFP) and (iv) EnvA protein (mCherry). Colocalization of the two channels is seen in (v) the starter cell. Scale bar in B (i and ii) is 250 µm. Scale bar in panel B (vi) is 50 µm. (C) Quantification of VIP starter cells in the MEC represented as the number of starter cells across the different layers of the MEC. (D) Quantification of input cells within different layers of the MEC relative to the total number of starter cells. (E) Representative image of monosynaptic input cells from different layers of the MEC, secondary visual cortex, perirhinal cortex, and the hippocampal formation. (Scale bar, 250 µm.) (F) Representative image of inputs arising from the CA1 region of the hippocampus, subicular complex, and retrosplenial cortex. (Scale bar, 250 µm.) (G) Input cell quantification for cells projecting from the CA1 and subicular complex (H) The total fraction of inputs from different brain regions projecting to VIP cells in the MEC. The total fraction of inputs is defined as the percentage of inputs a brain region sends to starter cells. (I) Representative images of the TVA-eGFP virus injected in VIPTdTom mice to confirm that the TVA-eGFP virus is expressed in VIP interneurons. (J) Representative images of the TVA-eGFP virus injected in the VIP^cre-negative animal show no viral expression leakage. (K) Quantification of colocalization between TVA-eGFP cell bodies and VIPTdTom to confirm specificity of virus in VIP^cre animals. (L) Input cell quantification for different brain regions in the isocortex, cortical subplate, and thalamus, as defined by the Allen Brain Atlas. The closed circle indicates sections that were sliced in the horizontal plane. (M) Representative images of inputs arising from the medial septum (MS). Scale bar for the image in the larger panel is 500 µm and for the inset in the yellow box is 100 µm. The CI for all graphs was calculated as a ratio of the number of inputs from a given brain region over the number of starter cells quantified in an animal. VIP^cre animals n = 4. In panels C, D, G, and L, full circles indicate sections sliced in the horizontal plane, and empty circles indicate sections sliced in the sagittal plane.

Fig. 2.

Interneurons in the MEC receive projections from the ADN. (A) Representative image of starter cells in PV^cre animals. Starter cells are highlighted in yellow and indicated by white arrowheads. Scale bar for the image in the larger panel is 500 µm and for the inset in the yellow box is 50 µm. (B) Representative image displaying input cells in the ADN and RSC of PV^cre mice injected with rabies-mediated retrograde virus. Scale bar for the image in the larger panel is 500 µm and for the inset in the yellow box is 50 µm. (C) Representative image of starter cells in SOM^cre animals. Starter cells are highlighted in yellow and indicated by white arrowheads. Scale bar for the image in the larger panel is 500 µm and for the inset in the yellow box is 50 µm. (D) Representative image displaying input cells in the ADN SOM^cre mice injected with the rabies-mediated retrograde virus. Scale bar for the image in the larger panel is 500 µm and for the inset in the yellow box is 50 µm. (E) Representative image displaying input cells in the ADN VIP^cre mice injected with the rabies-mediated retrograde virus. Scale bar for the image in the larger panel is 500 µm and for the inset in the yellow box is 50 µm. (F) Heat maps illustrate the quantification of the inputs to VIP, PV, and SOM MEC interneurons based on the convergence index for each cell line: VIP^cre (n = 4 mice), SOM^cre (n = 4 mice), and PV^cre (n = 2 mice). (G) Quantification of starter cells in the MEC for different interneuron populations. The closed circle indicates sections that were sliced in the horizontal plane. ADN—Anterodorsal thalamic nucleus, MS—Medial Septum, HDB—Horizontal Diagonal Band of Broca, and RSC—Retrosplenial cortex. VIP^cre animals n = 4, SOM^cre n = 4, PV^cre n = 2. In all panels full circles indicate sections sliced in the horizontal plane, and empty circles indicate sections sliced in the sagittal plane.

Thalamic ADN Projections Target MEC Interneurons, With a Preference for VIP Cells.

Direct ADN–MEC connections were reported to be very sparse (33) compared to projections to adjacent structures (34, 35), leading to the consensus that HD information predominantly arrives at the MEC via the pre- and parasubiculum. To determine the extent to which the previously observed connectivity pattern from our experiments in Fig. 1 is specific to VIP cells, the same experimental approach was applied to localize thalamic neurons projecting onto other MEC interneurons, such as PV (Fig. 2 A and B and SI Appendix, Fig. S1B) and SOM (Fig. 2 C and D and SI Appendix, Fig. S1C). Strikingly, ADN’s projection showed a strong preference toward VIP over SOM and PV cells (Fig. 2F, SI Appendix, Table S1, CI = 1.68 ± 0.68, CI = 0.29 ± 0.04 and CI = 0.26 ± 0.15, respectively). PV^cre line showed a higher number of starter cells (Fig. 2G, 122.5 ± 42.5), followed by SOM^cre (Fig. 2G, 116.25 ± 28.27) and VIP^cre lines (Fig. 2G, 65.75 ± 15.62). This distribution is consistent with other studies showing that PV is the largest medial entorhinal cortical interneuron subtype (36, 37), and demonstrates that the size of the starter cell population does not drive ADN inputs’ preferences for VIP cells. Following this, the connectivity pattern of the two main classes of excitatory superficial cells, the stellate (Fig. 3 A, B, and F and SI Appendix, Fig. S2) and the pyramidal cell populations (Fig. 3 C, D, and F and SI Appendix, Fig. S2), was also investigated. Remarkably, ADN cells targeting entorhinal cortical pyramidal and stellate cells were negligible or absent (Fig. 3 B and D and SI Appendix, Fig. S2 and Table S1, pyramidal cells: starter cells = 110.25 ± 25.91, CI = 0.02 ± 0.007; stellate cells: starter cells = 62 ± 25.66, CI = null). Thus, the sparsity of ADN projections onto MEC might result from their selectivity for entorhinal inhibitory cells. Although VIP cells in the MEC represent about 11% of GABAergic cells in the MEC (22), this result suggests that the VIP cell population might be the principal direct relay of anterodorsal thalamic HD signal to the MEC.

Fig. 3.

Tracing of inputs to excitatory cells in the MEC. (A) Representative image of starter cells in sim1^cre animals. Starter cells are highlighted in yellow and indicated by white arrowheads. Scale bar for the image in the larger panel is 500 µm and for the inset in the yellow box is 50 µm. (B) Representative image of the absence of input cells in the ADN and RSC of sim1^cre mice injected with rabies-mediated retrograde virus. Scale bar for the image in the larger panel is 500 µm and for the inset in the yellow box is 50 µm. (C) Representative images of starter cells in pOx^cre animals. Starter cells are highlighted in yellow and indicated by white arrowheads. Scale bar for the image in the larger panel is 500 µm and for the inset in the yellow box is 50 µm. (D) Representative image of the absence of input cells in the RSC of pOx^cre mice injected with rabies-mediated retrograde virus. Few input cells could be identified in the ADN of pOx^cre mice. Scale bar for the image in the larger panel is 500 µm and for the inset in the yellow box is 50 µm. (E) Representative image of input cells in the MS/HDB region in pOx^cre animals. Scale bar in the image is 500 µm. (F) Quantification of starter cells in the MEC for different excitatory cell populations. Full circles indicate sections sliced in the horizontal plane, and empty circles indicate sections sliced in the sagittal plane. (G) Heat maps depict input tracing to sim1^cre and pOx^cre MEC excitatory cells. Each row displays the convergence index for each cell line: sim1^cre (n = 3 mice) and pOx^cre (n = 4 mice). ADN—Anterodorsal thalamic nucleus, MS—Medial Septum, HDB—Horizontal Diagonal Band of Broca, and RSC—Retrosplenial cortex.

VIP Cells Receive More Intra-and Interregional Projections Than the Other Medial Entorhinal Cellular Subtypes.

Within multiple neuronal subtypes in the MEC, VIP cells appeared to receive the densest projections from other entorhinal cortical cells (Fig. 2F and SI Appendix, Fig. S1A and Table S1, VIP CI = 17.33 ± 6.57, PV CI = 3.62 ± 0.71; SOM CI = 3.36 ± 0.60; Fig. 3G and SI Appendix, Fig. S2, pyramidal cells CI = 2.33 ± 0.92; stellate cells CI = 5.49 ± 1.88). We observed similar projection patterns from CA1 (Fig. 2F and SI Appendix, Fig. S2; VIP CI = 8.08 ± 4.62, PV CI = 1.68 ± 0.50, SOM CI = 1.65 ± 0.41; Fig. 3G and SI Appendix, Fig. S2; pyramidal cells CI = 0.33 ± 0.09; stellate cells: 0.01 ± 0.01), the subicular complex (Fig. 2F and SI Appendix, Fig. S2; VIP CI = 5.22 ± 1.55, PV CI = 2.76 ± 0.39, SOM CI = 2.27 ± 1.48; Fig. 3G and SI Appendix, Fig. S2; pyramidal cells CI = 1.03 ± 0.34, stellate cells CI = 1.81 ± 0.75), and the retrosplenial cortex (RSC) (Fig. 2F and SI Appendix, Fig. S2; VIP CI = 1.01 ± 0.51, SOM CI = 0.04 ± 0.02, PV: null; Fig. 3G and SI Appendix, Fig. S2; pyramidal and stellate cells: null). Inputs from the medial septum/horizontal diagonal band contacted all cell types, except stellate cells (Figs. 2F and 3 E and G and SI Appendix, Fig. S2), with a slight preference for VIP cells (Fig. 2F and SI Appendix, Fig. S2, VIP CI = 0.38 ± 0.05; PV CI = 0.07 ± 0.0; SOM CI = 0.33 ± 0.29; Fig. 3G and SI Appendix, Fig. S1 pyramidal cells CI = 0.01 ± 0.009; stellate cells: null). Perirhinal cortex inputs to MEC cells were also very sparse and restricted to VIP and stellate cells, the latter receiving the most inputs (SI Appendix, Fig. S1, VIP CI = 0.028 ± 0.01, SOM CI = null, PV: null, SI Appendix, Fig. S1, stellate cells CI = 0.34 ± 0.20, pyramidal cells: null). Regarding the LEC, no input cells were found in any interneuron mouse lines, but some inputs projecting onto pyramidal and stellate cells were spotted (SI Appendix, Fig. S3). VIP interneurons are known to inhibit other interneuron populations, such as PV and SOM cells, in various cortical regions. To investigate whether VIP cells provide presynaptic control over PV and SOM cells in the MEC, we quantified the overlap between VIP-immunopositive cells and rabies-labeled input cells in PV- and SOM-injected animals. Immunofluorescence analysis was performed in these animals (SI Appendix, Fig. S4). In the PV^cre animals, the average number of double-labeled cells with rabies was 2.5 ± 0.5 cells (SI Appendix, Fig. S4C), corresponding to 2.5% of input cells labeled for VIP. In SOM^cre animals, 6.5 ± 3.6 VIP-positive cells were colabeled with rabies (SI Appendix, Fig. S4C) representing 3.84% of input cells. We also quantified colocalization between input-labeled cells and Calretinin (CR) neurons in both PV^cre and SOM^cre animals. In SOM animals, 6.5 ± 2.5 cells were double-labeled (SI Appendix, Fig. S4C). This corresponds to 8.33% of input cells comprising 3.76% of the total unique cell population (233.5 cells). These results support the notion that input-receiving cells for SOM cells in the MEC include CR+ interneurons, many of which likely overlap with VIP+ interneurons. No overlap between CR and input cells was observed in PV Cre animals (SI Appendix, Fig. S4E).

VIP Cells Are Activated by Spatial Changes in the Environment and Are Essential for Spatial Memory.

As MEC VIP cells are densely connected with key memory and space-coding brain regions, we proceeded to assess the putative role of VIP cells in spatial memory (Fig. 4A). VIPtdTom mice were assessed on the MEC-dependent novel object location (NOL) task (11, 38–40). As previously reported, mice could discriminate an object’s new location in the dissimilar condition but not in the extrasimilar condition (Fig. 4B, dissimilar: 0.33 ± 0.04, one-sample t test: t = 0.414, df = 6, P < 0.0001; extrasimilar: −0.02 ± 0.05, one-sample t test: t = 8.265, df = 8, P = 0.6933; t test dissimilar vs. extrasimilar: t = 5.484, df = 14, P = 0.0002). Differences in discrimination ratio were not driven by exploration time during the sampling or the test (Fig. 4C, exploration time—sampling control: 35.82 ± 4.63, extrasimilar: 34.23 ± 2.41; dissimilar: 33.51 ± 2.31, one-way ANOVA F(2, 19) = 0.1441, P = 0.8667; exploration time—test control: 13.86 ± 1.61, extrasimilar: 13.53 ± 1.09, dissimilar: 15.89 ± 1.31, one-way ANOVA F(2, 19) = 0.99, P = 0.39; and SI Appendix, Fig S5 B and C) or sex (SI Appendix, Fig. S5A). To determine whether VIP cells were activated during the NOL task, we imaged the expression of the c-Fos immediate early gene protein (41) 90 min after the test phase (Fig. 4 A and D). No difference in the c-Fos population was found between control and test conditions when looking at the total number of c-Fos+ cells in the MEC (Fig. 4E, behavioral control: 13,132 ± 2041; dissimilar: 17,214 ± 1813; extrasimilar: 18,214 ± 1,490; one-way ANOVA with Tukey’s comparison test F(2, 13) = 2.246, P = 0.1453). However, the number of VIP cells expressing c-Fos (VIP/c-Fos+ cells) increased when animals were engaged in a discrimination-based task (Fig. 4F, behavior control 13 ± 1.17, dissimilar 24.49 ± 3.23, extrasimilar 27.65 ± 1.78; one-way ANOVA with the Tukey comparison test: F(2, 13) = 13.78, P = 0.0006, behavior control vs. extrasimilar: P = 0.0008, behavior control vs. dissimilar: P = 0.0052), but independently of the behavioral conditions (extrasimilar vs. dissimilar: p = 0.58). When parsing it through the layers, the superficial layers showed a greater level of activation for the total c-Fos population only (SI Appendix, Fig. S6A, two-way ANOVA with Tukey’s multiple comparisons test: layer F(1, 26) = 93.50, P < 0.0001; behavior F(2, 26) = 2.473, P = 0.1039; interaction F(2, 26) = 1.817, P = 0.1825; superficial vs. deep: behavior control: P = 0.0001, dissimilar: P < 0.0001, extrasimilar: P < 0.0001; superficial layers: behavior control vs. dissimilar: P = 0.1515, behavior control vs. extrasimilar: P = 0.0243, dissimilar vs. extrasimilar: P = 0.6771; deep layers: behavior control vs. dissimilar: P = 0.9231, behavior control vs. extrasimilar: P = 0.9891, dissimilar vs. extrasimilar: P = 0.9712), as no layer differences were observed when focusing on the VIP cells only (SI Appendix, Fig. S6B, two-way ANOVA with Tukey’s multiple comparisons test: layer F(1, 26) = 0.1582, P = 0.6941; behavior F(2, 26) = 4.583, P = 0.0197; interaction: F(2, 26) = 0.6008, P = 0.5558; superficial vs. deep: behavior control: P = 0.8025, dissimilar: P = 0.5832, extrasimilar: p = 0.3291; superficial layers: behavior control vs. dissimilar: P = 0.4225, behavior control vs. extrasimilar: P = 0.2104, dissimilar vs. extrasimilar: P = 0.8969; deep layers: behavior control vs. dissimilar: P = 0.8939, behavior control vs. extrasimilar: P = 0.0448, dissimilar vs. extrasimilar: P = 0.1334). Altogether, these results demonstrate that spatial memory is associated with the activation of VIP cells, independently of behavioral performance.

Fig. 4.

VIP interneuron recruitment is sensitive to changes in spatial environment and is necessary for spatial memory. (A) Schematic of the novel object location (NOL) task. (B) VIP^cre mice can significantly discriminate a new object’s location in the dissimilar, but not the extrasimilar, version of the NOL task (extrasimilar n = 7, dissimilar n = 9, ***P < 0.001). (C) VIP^cre mice in the behavior control (n = 6), extrasimilar (n = 7) and dissimilar conditions n = 9) spend similar time exploring objects during the sampling phase and test phase of the paradigm. (D) (i) Representative image of c-Fos immunostaining in the MEC sections from VIPtdTom mice. In panels (ii and iii), colocalized VIPtdTom+ and c-Fos+/GFP-expressing cells are indicated in white. Panel (iv) shows the merged image to visualize colocalization. Scale bar in panel (i) is 500 µm and in panels (iv) is 100 µm. Note that the images in panels (ii–iv) are from a different animal than the one shown in panel (i). (E) Quantification of the c-Fos+ cells population across all MEC layers. Behavior control n = 6, dissimilar n = 5, extra similar n = 5. (F) Quantification of the VIP/c-Fos+ cells across all the MEC layers. Behavior control n = 6, dissimilar n = 5, extra similar n = 5. (**P < 0.01; ***P < 0.001) (G) The inhibitory AAV-DIO-hSyn-hM4Di-mCherry virus was injected unilaterally into the MEC of VIP^cre mice. Injection spread is visualized in panel (i). In panel (ii), hM4Di-mCherry-expressing cell bodies are marked in white. Scale bar in panel (i) is 500 µm and in panel (ii) is 100 µm. Panel (iii) shows the quantification of injection spread across the superficial (LI-III) and deep layers (IV–VI) of the MEC. N = 6 animals tested in the dissimilar condition. (H) JHU (1 mg/kg) or saline was administered 30 min before the sampling phase. (I) No difference in the exploration time during the sampling (Left panel) and test (Right panel) was observed between the control (n = 7) and JHU-administered group (n = 6). (J) Mice that received JHU prior to sampling show a significant impairment in their ability to discriminate an object’s new location (***P < 0.001). All data are represented with ± SEM.

To elucidate the role of VIP cells in spatial memory, VIP^cre mice were injected with the chemogenetic viral construct AAV8.hSyn.DIO.hM4Di.mCherry virus and later were injected with JHU (42) to drive the selective inhibition of VIP cells (Fig. 4 G and H). While our injections targeted all layers of the MEC, most hM4Di cell bodies were found in the superficial layers (Fig. 4 G, iii, LI-III: 70.67 ± 10.77; LIV-VI: 23.17 ± 10.38). Two control groups were used: the JHU control group did not receive a viral injection but did receive JHU before behavior, while the virus control group was transfected with the AAV8.hSyn.DIO.hM4Di.mCherry virus but instead of the JHU drug, received saline, before sampling. Despite a slight but significant increase in the exploration time in the virus control group compared to the JHU control group (SI Appendix, Fig. S4D, unpaired t test virus control vs. JHU control: t = 2.451, df = 6, P = 0.0497), no difference in the behavioral performance was detected between groups (SI Appendix, Fig. S4C, unpaired t test virus control vs. JHU control: t = 0.752, df = 6, P = 0.4803). Therefore, both groups were pooled together and referred to as the behavioral control group (Fig. 4 I and J). As JHU was injected 30 min before the sampling, VIP cell activity was reduced throughout the behavioral procedure. Although JHU injection did not entirely abolish the ability of mice to detect an object’s new location in the dissimilar condition (Fig. 4J, one-sample t test behavior control: t = 10.35, df = 7, 0.24 ± 0.02, P < 0.0001; dissimilar: t = 3.394, df = 5, 0.08 ± 0.028, P = 0.0194), JHU severely altered the ability of mice to perform the task (Fig. 4J, unpaired t test behavior control vs. JHU: t = 4.889, df = 12, P = 0.0004). This effect could not be explained by changes in the exploratory behavior, as the time spent exploring objects was similar between groups for both the sampling (Fig. 4I, behavior control: 35.16 ± 21.37, JHU: 34.21 ± 39; unpaired t test behavior control vs. JHU: t = 0.23, df = 2, P = 0.8220) and the test (behavior control: 14.26 ± 15.91, JHU 18.20 ± 16.68, unpaired t test behavior control vs. JHU: t = 1.682, df = 12, P = 0.1184). Similar results were observed when mice received bilateral virus injection (SI Appendix, Fig. S4 F and G, exploration time sampling, unilateral: 34.21 ± 39, bilateral: 38.10 ± 80.39, unpaired t test unilateral vs. bilateral: t = 0.485, df = 8, P = 0.6403; unpaired t test: discrimination ratio, unilateral: 0.08 ± 0.02, bilateral: 0.01 ± 0.09, t = 0.942, df = 8, P = 0.7619). Hence, this result demonstrates that the inactivation of MEC VIP cells impairs discrimination between spatial memories that share common features.

Discussion

Most of our understanding of the biological mechanisms supporting MEC’s ability to create spatial representations comes from the study of principal cells. However, the role of interneurons in shaping spatial representations in the MEC remains limited. Our study corroborates results obtained in various brain regions, supporting the idea that VIP cells are important to relay information provided by long-range afferents into the local network, as observed in other brain regions such as the amygdala (24), anterior insular cortex (32), and the prefrontal cortex (31). More specifically, this report sheds light on the critical position of VIP cells in MEC connectivity with regions known to play central roles in spatial representation and memory. First, we demonstrated that MEC interneurons, particularly those that are VIP-expressing, are densely connected not only to the local circuitry but also to the hippocampus, the medial septum, the subicular complex, and the retrosplenial cortex. Most surprisingly, retrograde tracing analysis unveiled the existence of a direct thalamocortical pathway in which the ADN sends direct projections specifically to the MEC inhibitory network. Importantly, we observed that VIP cells are more densely connected to thalamic inputs, suggesting that these cells play a key role in relaying a thalamic HD code to other MEC cells. Second, we demonstrated that VIP cells were activated in response to spatial environmental changes, suggesting a potential role for these cells in sensing environmental alterations. Finally, we found that chemogenetic inhibition of VIP cells profoundly impaired the ability of mice to discriminate between memories that share common features.

ADN Projections to the MEC Are Selective to the Inhibitory Network and Essentially Target VIP Cells.

This study reports a unique selective projection from the ADN to MEC interneurons, where most anterodorsal thalamic inputs target VIP cells. Such an observation challenges the popular view on the propagation of the HD signal from the thalamus to cortical regions, wherein the MEC is thought to mostly receive the thalamic HD signal indirectly through superficial layers of the presubiculum (33, 43–46). Nevertheless, the current computational models of the HD system mostly overlook the contribution of this pathway, instead highlighting the MEC-projecting pyramidal cells in the presubiculum that receive direct projections from the AND (43–45) as the source of the HD signal in the MEC. Hence, our study raises several questions: 1) Do VIP cells receiving thalamic HD signal modulate the same population of medial cortical HD cells as the presubiculum or a different subpopulation? 2) Are these two pathways activated in the same context, or do they depend on environmental cues and features? 3) How do these two pathways modulate HD computation in the MEC? Interestingly, a recent study from Nassar et al. (47) similarly reported direct projections from the anterior thalamus to VIP cells in the presubiculum, a region where most of the work investigating the role of the inhibitory drive in the HD signal has been performed (43–45). Together with our report, this recent work suggests that VIP cells might be a common relay of thalamic inputs into space-coding cortical regions, thus shaping space representation in the MEC. Indeed, it is interesting that neither inhibiting MEC PV nor SOM cells affected medial entorhinal HD cells (17). What could be the computational role of an ADN to MEC VIP projection? We propose that this pathway could provide context-dependent gain modulation of HD representations in the MEC. Specifically, VIP-mediated disinhibition might act as a gating mechanism that determines whether incoming thalamic HD signals influence cortical attractor dynamics—an idea that parallels network gain control models in the thalamus and HD system more broadly (48). In such a framework, VIP activity could bias the realignment of grid cell manifolds following changes in environmental landmarks (such as the change in the object’s location in our task) or spatial uncertainty. This mechanism would allow the MEC to flexibly switch between internally maintained and externally reset spatial codes. Conceptually, this VIP relay may serve not simply to pass on HD signals, but to selectively permit their influence, based on environmental cues or novelty, thereby enabling adaptive updates to the animal’s internal map. Given our retrograde tracing data, it is likely that MEC VIP cells receive input from local excitatory cells, CA1, and the subiculum. This suggests that MEC VIP cells may also receive inputs from many space-encoding cells, such as place cells or grid cells. Therefore, future investigations probing the contribution of the ADN–MEC pathway in space encoding will be important.

Finally, the number of thalamic afferents reaching a cell population is not the sole determinant of a pathway’s importance for target populations. Although ADN innervation to PV and SOM interneurons is sparse compared to the VIP cell population, the weight of this connection can be balanced by cellular and molecular plasticity properties, facilitating or depressing the integration of thalamic information. To address this question, future work involving the recording of interneurons in physiological conditions during manipulation of ADN–MEC projections would be necessary to determine how the thalamic HD signal influences entorhinal cortical interneurons. Interestingly, a previous study (49) has shown that connections exist from the MEC to the Anteroventral Nucleus (AVN) of the thalamus, while direct projections from the MEC to the ADN were not observed. While the nature of this feedback projection is unknown, further investigation into the connectivity between the MEC and various thalamic nuclei will be necessary to better understand the circuitry between these two structures.

VIP Cells in Memory: A Ubiquitous Mechanism for Memory Formation.

Beyond the MEC, spatial memory encoding also depends on structures such as the CA1, subiculum, RSC, and ADN. The RSC, for example, has been implicated in object location processing (50) and may also contribute to object identity coding via tactile signals. In particular, Lande et al. (51) demonstrated that neurons in the agranular RSC represent the spatial location at which whiskers contact objects, underscoring a role in somatosensory-driven spatial encoding. Furthermore, the ADN plays a crucial role in multiple aspects of spatial processing; lesions to this structure impair allocentric place learning, direction learning, path integration, and even certain nonspatial functions such as temporal discrimination (52). In parallel, studies have shown the presence of object-vector coding in the MEC (10) and the importance of contralateral MEC projections for these cells (53). Given that VIP interneurons receive a majority of their input from the regions mentioned above, they present a compelling link between sensory-derived object information and spatial representations. These multiple inputs position VIP cells as potential modulators of the diverse components involved in spatial memory.

VIP cells have been investigated due to their ability to release principal cells from ongoing inhibition, which is important for associative memory (24, 26). Indeed, in the MEC, approximately two-thirds of the VIP cells coexpress calretinin (22), a marker for disinhibitory interneurons (54), in a proportion that is relatively similar to what is observed in the presubiculum (47). Our findings on the effect of silencing MEC VIP cells in a spatial discrimination task further validate the significance of this cell group in memory processes. However, results obtained from c-Fos imaging suggest that VIP cells may have a more dynamic function, in addition to their role in associative memory. Specifically, the recruitment of VIP cells increased when changes were made to the spatial layout of objects, independent of any corresponding changes in behavioral response, suggesting a role of VIP cells in sensing changes in the environment. This observation aligns with recent studies examining the mechanisms of discrimination in the hippocampus. In a virtual environment where mice learn the location of a reward, small changes in the environment did not change the behavioral response of the animal, although DG’s cell activity changed upon exposure to a new, yet very similar, environment (55). Interestingly, when the new environment was made more dissimilar, mice could learn the new location of the reward, with DG and CA1 areas showing altered activity in response to the dissimilar environment (55). The result suggests that separate representations of the two environments were implemented throughout the hippocampus. This work demonstrated that 1) neural discrimination does not parallel behavioral discrimination and 2) novelty leads to an adaptation in the behavioral response only if environmental changes are salient enough to modulate the activity of specific brain regions, such as CA1 (55), to allow the formation of a representation of new experiences. A similar observation has been made in the prefrontal cortex, where ventral hippocampal inputs to prefrontal VIP-expressing cells drive different representations of the open and closed arms of an elevated-plus maze (31). Here, more VIP cells were activated after the extra similar and dissimilar conditions, compared to the control condition in which none of the objects were moved. Hence, MEC VIP cells may receive a novelty signal, enabling the activation of different populations of excitatory cells within the MEC. Specifically, stellate cells, which form synapses with both DG and CA3 cells (11, 56) and are essential for recognizing novel object location (11), emerge as potential targets for VIP-mediated disinhibition. Hypothetically, if VIP interneurons participate in the disinhibition of grid cells, one of the functions in part supported by stellate cells (57), this could lead to a greater strength of these neurons firing into downstream hippocampal cells, increasing the excitability of these neurons. Moreover, a portion of stellate cells have been shown to express HD activity (57), known to exert a strong influence on grid cell activity (58, 59). While our results point to a putative role played by VIP cells, the specific connectivity pattern between VIP and stellate cells remains to be determined.

More detailed investigations are needed to reconcile the plausible role of VIP cells in the modulation of various spatial inputs, including cortical HD signals, and their implication in spatial memory. Determining 1) how inhibition of cortical VIP cells interferes with memory and 2) if and through which connectivity scheme cortical VIP cells control HD cells’ activity in the MEC may help in elucidating the functional role of this cell population in the spatial memory circuitry.

Technical Limitations.

Several technical limitations must be considered, as they might influence the results reported in this study. First, although many entorhinal cortical cell types were studied, our study covered selected cellular subtypes. In particular, excitatory cells from MEC deep layers were not investigated here. Therefore, it is essential to consider that some studied regions potentially send most of their inputs to MEC deep layers, as it is already known for the retrosplenial cortex (60). Second, using rabies-mediated retrograde labeling unilaterally, we found few input cells in the contralateral hemisphere. Thus, some anatomical pathways might have been overlooked, such as those involving stellate and pyramidal cells, which are typically known to receive substantial projections from the contralateral hemisphere (60, 61). Third, the number of starter cells for each cellular subtype studied here seems to reflect the proportion of cellular subtypes that are commonly observed in cortical structures. However, as such, we cannot rule out that quantitative differences between input cells are due to transfection efficacy that might vary between the different cre lines used here.

Conclusion

This study corroborates similar results obtained in various brain regions, supporting the idea that VIP cells are important to relay information provided by long-range inputs into the local network. More specifically, this study reveals the existence of a direct thalamocortical pathway to the MEC through VIP cells that could support essential computations in the HD system. Importantly, we demonstrate that this cell population is involved in spatial memory and could be essential for detecting environmental changes. Altogether, this work sheds light on the traditionally understudied MEC VIP cells and provides insights into their contribution to MEC function, spatial cognition, and memory.

Materials and Methods

Animals.

This study was carried out following the recommendations of the Canadian Council on Animal Care and the McGill University Animal Care Committee. The McGill University Animal Care Committee approved the protocol. Animals were housed in a temperature-controlled room with a 12/12 h light/dark cycle and food and water ad libitum. Homozygous Vip-IRES-cre (VIP^cre, #010908, The Jackson Laboratory) mice were crossed with homozygous Ai9 lox-stop-lox-tdTomato cre-reporter mice (#007905, The Jackson Laboratory) to generate VIP-tdTom mice in which tdTom (Tom) is exclusively expressed in cells that have cre recombinase. Dr. Matthew Nolan, University of Edinburgh, kindly gave sim1^cre mice (C57Bl6/J). Sst-IRES-cre (SOM^cre, #018973), Pvalb-IRES-cre (PV^cre, #017320), and pOxr1-cre (pOx^cre, #030484) were purchased from The Jackson Laboratory.

Stereotaxic Surgeries for Viral Transfection.

Five- to eight-week-old mice were anesthetized, their body temperature was maintained with a heating pad, and their eyes were hydrated with gel (Optixcare). Carprofen (2 mg/g) and saline (0.5 mL) were administered subcutaneously during the surgery. For each craniotomy, two injection sites were used. For the first site, the injection pipette was lowered to 3.65 mm from the bregma; for the 2nd site, it was lowered to 3.45 mm from the bregma. The pipette was moved 0.280 mm anterior to the transverse sinus for both injection sites to target the MEC. The injection depth was determined by the extent to which the micropipette bent once it hit the dura lining surface of the brain. Approximate depths for 5-wk-old mice were 1st injection: −2.240, 2nd injection site: −3.00. For each injection site, the pipette was retracted 200 µm after the first bend, and the virus was injected at three different depths along the dorsal–ventral axis at 300 µm intervals along the extent of the MEC, such that the last injection site usually corresponded to −1.450 to −1.750 along the DV axis. After each injection, the micropipette was left undisturbed for 3 to 4 min to allow for the infusion of the virus before injecting into the next site. A total of 400 µL of virus was injected into each animal.

Injection timeline for rabies-mediated viral tracing.

VIP^cre (n = 4), sim1^cre (n = 3), pOx^cre (n = 4), PV^cre (n = 2), and SOM^cre (n = 4) mice were used for tracing monosynaptic inputs to MEC. These animals were first unilaterally injected with a 1:1 combination of AAV2/8-hSyn-FLEX-TVA-P2A-eGFP-2A-oG (ULaval Vector Core). 4 wk later, animals were anesthetized as previously described, and RABV-DeltaG-EnvA-mCherry virus was injected using scar on the skull as a guide for location. One week postinjection, animals were perfused (4% PFA) to detect input and starter cells.

Injection timeline for chemogenetic viral expression.

For chemogenetic experiments, (n = 15) VIP^cre mice were injected bilaterally or unilaterally with the cre-mediated Designer Receptor Exclusively Activated by Designer Drugs (DREADDs) viral construct AAV8.hSyn.DIO.hM4Di.mCherry (Addgene, 4.3e12 GC-ml, diluted to a third in sterile PBS) using the injection strategy as described above.

Habituation and Handling for Behavior Experiment.

c-Fos experiments.

One week before testing in the novel object location paradigm, VIPtdtom (n = 14) mice were handled daily for 3 to 5 min for five consecutive days.

Chemogenetic experiments.

Animals were subjected to behavioral testing only after recovering from surgeries (2 wk), followed by two weeks of handling. For the first week, animals were handled for 3 to 5 min. For the second week, they were restrained by the scruff, and the animals were administered i.p. injections with saline every couple of days.

Novel Object Location Paradigm.

To test spatial memory, we adapted a novel object location paradigm from ref. 39. Briefly, mice were placed in an opaque circular open field (40 cm in diameter) surrounded by black walls with salient high-contrast distal cues. During the four days of habituation, animals were placed daily for 10 min in the open field. During the last two days of habituation, a set of 2 new and different objects was introduced in the open field. On day 5, animals were placed for 15 min in the open field (sampling) and reintroduced 90 min later for 6 min (test). During the sampling, two identical objects were aligned horizontally in the center of the open field. During the test, one object was moved to a novel location (dissimilar: 12 cm from the original location; extrasimilar: 3 cm from the original location) while the other remained in its original location. The location and choice of the objects were counterbalanced between different groups. For both tasks, the time spent exploring each object in the sample and test phases was measured, and animals that did not explore at least 10 s for each object during the sampling phase were excluded from the analysis. In the control group, the two objects were left unmoved during the test phase. The discrimination ratio was calculated as follows: Exploration time_moved– Exploration time_unmoved)/ (Exploration time_moved+ Exploration time_unmoved) (s).

Two experimenters scored the exploration times manually, and the average of their exploration times was used for analysis.

c-Fos experiment.

VIPtdTom animals were used and perfused (4% PFA) 90 mins after the test session, and their brains were serially sectioned for c-Fos staining.

Chemogenetic experiments.

VIP^cre mice were injected with AAV8.hSyn.DIO.hM4Di.mCherry virus in the MEC. VIP^cre mice in the control group did not receive any viral injection. JHU37160 [JHU, HelloBio #HB6261, 1 mg/kg, diluted in saline (42)] or saline was injected 30 mins before sampling. Each animal was tested twice, once with saline and once with JHU 2 wk apart. Mice were then perfused (4%PFA) after testing, and the brains were sectioned to confirm viral expression and specificity of the injection location.

Immunohistochemistry.

Mice were deeply anesthetized and intracardially perfused with 4% PFA in PBS. Fixed brains were snap-frozen in dry ice-chilled isopentane before being cut into 35-μm-thick sections using a cryostat (Leica Microsystems).

Rabies injections.

For circuit mapping experiments, whole brains from injected animals were sliced in the horizontal and sagittal plane to retrieve whole brain sections. Postslicing, the sections were rinsed in 1X PBS (3 × 5 min). For SI Appendix, Fig. S4, the VIP and CR immunostaining protocols were similar to those described in our previous study (22). Briefly, primary antibodies were used at the following concentrations: VIP—1:500 (rabbit anti-VIP, ImmunoStar 722001/20077, ImmunoStar, Dietzenbach, Germany) and CR—1:3,000 (rabbit anti-Calretinin, Swant 7697, Marly, Switzerland) for 48 h in a rotating shaker at 4 °C. Following primary antibody incubation, slices were incubated with secondary antibodies at 1:1,000 dilution: Alexa Fluor 647 (ThermoFisher Scientific) for VIP staining and Alexa Fluor 350 (ThermoFisher Scientific) for CR staining for 2 h at RT.

c-Fos immunostaining.

The sections were rinsed in 1X PBS (3x10 min) and then incubated in 1X PBS containing 0.3% Triton-X with the c-Fos antibody (1:1,000, Abcam AB222699) at room temperature for 24 h. After one rinse in 1X PBS, the slices were incubated with the secondary antibody (1:1,000, ThermoFisher Scientific #A21206, Alexa Fluor 488) for 2 h at RT and finally rinsed 3 × 10 min in 1X PBS. DAPI mounting medium was used as a counterstaining.

Tracing of Input Regions for Rabies Circuit Tracing Experiments.

For visualization, analysis, and quantification purposes, the sections from the rabies experiment were imaged with a VS-120 Olympus slide scanner under epifluorescence in DAPI, FITC, and TRITC channels using an optical filter set from Semrock (Part Number: DA/FI/TR 4X4M-C-000), an Orca r2 Hamamatsu monochrome camera, and an Olympus ×10 objective (Molecular and Cellular Microscopy Platform—Douglas Research Center, Montreal, Canada). Virtual Slide Images (.vsi) files acquired from the slide scanner were opened in FIJI using the BIOP VSI Reader (EPFL, Lausanne, Switzerland) and exported as individual TIFF files for further analysis in FIJI (Image J version 2.1.0) and QuPath (Software version 0.4.3.1).

Mouse brain atlas and regional analysis.

The correct localization of input cells (mCherry+) in the MEC was validated using ABBA (62) [Aligning Big Brain Atlas—EPFL, Lausanne (https://biop.github.io/ijp-imagetoatlas/)] to register and align the Allen Atlas using both affine and spline transformation on sagittal and horizontal sections. The output was imported from ABBA, and manual counting was performed in the region of interest (ROI) outlined using the Allen Atlas in QuPath. The convergence index was defined as the total number of inputs targeting a single starter cell (Total number of inputs for an ROI/Total number of starter cells). The fraction of inputs was defined as the percentage of inputs a brain region sends to starter cells in the MEC (Total number of inputs for an ROI/Total number of presynaptic inputs in the entire brain) for each animal.

Image Acquisition and Data Analysis for c-Fos Quantification.

VIP, c-Fos, and VIP/c-Fos+ cells were quantified using the optical fractionator method in StereoInvestigator software with a Zeiss ApoTome structured illumination device on a widefield microscope. Contours were drawn to delineate the MEC’s superficial (L1-III) and deep layers (LIV-VI) using DAPI as a reference. Using live counting, an experimenter blind to the experimental conditions counted all markers according to optical dissector inclusion–exclusion criteria at each cell’s widest point. Separate markers were used to count the number of Tom (VIP), c-Fos (GFP), and colocalized or double-labeled cells (yellow). For normalization, the total number of VIP/c-Fos+ cells were normalized to the total number of VIP cells present in an animal for each condition.

Data Analysis and Statistics.

Statistical analyses were performed with GraphPad Prism 8. Unpaired Student t tests, one-way ANOVA with Tukey multiple comparison corrections were used. Data are presented as mean ± SEM with individual data points. Differences were considered significant when P < 0.05. Additionally, column statistics against a theoretical mean of 0 were performed on the behavioral analysis; if α=0.05, the animal was considered to have learned during the paradigm.

Data Repository.

Images of the retrograde rabies experiments are available on the Federated Research Data Repository: (63).

Please note that the images are not annotated with the Allen Atlas.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Immunoflourescent Images data have been deposited in Federated Research Data Repository (https://www.frdr-dfdr.ca/repo/dataset/e92b7aa5-a5f3-4f38-99a2-1c6d1b225489) (63).