To determine whether vHPC-LS neurons played a causal role in allowing animals to discriminate between novel and familiar mice, we chemogenetically silenced vHPC-LS neurons while evaluating the effects on behavior in the social discrimination task (SDT). To selectively silence vHPC-LS neurons, we injected C57BL/6J mice in the LS with a retrogradely transporting Cre virus and an AAV5 virus expressing Cre-dependent inhibitory DREADD (hM4Di)(Armbruster et al., 2007) tagged with mCherry (or a control virus expressing only mCherry) in the vHPC (Figure 1A and B). Three weeks after surgery, we injected mice (i.p.) with either saline or CNO 30 min prior to the SDT (Figure 1C). In the SDT, mice were allowed to explore an arena containing age- and sex-matched novel and familiar conspecifics (Figure 1C). Typically, mice spend more time in the proximity of a novel conspecific relative to the familiar conspecific (Moy et al., 2004). We found that chemogenetic inhibition of vHPC-LS neurons by administration of CNO but not saline in hM4Di expressing mice (male and female) disrupted the ability of mice to preferentially spend more time in the proximity of the novel animal (Figure 1F; two-factor ANOVA with drug condition and conspecific identity as factors; interaction: p=8.4E-04, main effect of conspecific identity: p=0.003; post hoc Sidak multiple comparison tests: Saline: p=2E-05, CNO: p=0.971; Figure 1 - Figure Supplement 1). CNO administration in mCherry-expressing control mice did not affect their natural preference for the novel conspecific (Figure 1D; two-factor ANOVA with drug condition and conspecific identity as factors; interaction: p=0.638, main effect of conspecific identity: p=6.40E-06; post hoc Sidak multiple comparison tests: Saline: p=1.10E-04, CNO: p=1.90E-05). In both saline and CNO conditions, the discrimination scores of the mCherry mice show that mice preferentially spend more time investigating the novel conspecific relative to familiar conspecific (Figure 1E; one sample t-test, Saline: p=0.0103, CNO: p=0.0013) while inhibition of vHPC-LS neurons disrupts the preference for the novel mouse (Figure 1G; one sample t-test, Saline: p=0.0009, CNO: p=0.9222).

To determine if the effects observed with vHPC-LS inhibition are socially specific, we ran the mice through a food discrimination assay in which mice were allowed to explore an arena containing novel and familiar foods. Unlike with conspecifics, mice are neophobic and do not prefer novel foods (File, 2001). Therefore, we expected mice would spend less time around the novel food in comparison to the familiar food. All mice exhibited a preference for the familiar food regardless of virus or drug condition (Figure 1H–K; two-factor ANOVA with food identity and drug condition as factors; mCherry: interaction: p=0.254, main effect of food preference: p=0.003, post hoc Sidak multiple comparison tests: Saline: p=0.0323, CNO: p=9E-04; discrimination score: one sample t-test, Saline: p=0.0134, CNO: p=0.008; hM4Di: interaction: p=0.508, main effect of food preference: p=3.2E-05, post hoc Sidak multiple comparison tests: Saline: p=1.7E-04, CNO: p=0.003; discrimination score: one sample t-test, Saline: p=0.0016, CNO: p=0.0419). This finding suggests that the effects observed with vHPC-LS inhibition are socially specific.

To determine if the reduction in preference for a novel conspecific in the SDT could arise from anxiety-like behaviors, we ran mice through an open field assay. We found that chemogenetic silencing of vHPC-LS neurons had no effect on the time spent in the center of the arena (Figure 1L and N; paired t-test, mCherry: p=0.3167, hM4Di: p=0.1837). Additionally, chemogenetic inhibition of vHPC-LS neurons had no effect on locomotion, as evidenced by the absence of any changes in the velocity of the animals (Figure 1M and O; paired t-test, mCherry: p=0.2694, hM4Di: p=0.7886). Thus, ruling out the possibility that altered locomotion could be contributing to the ability of mice to discriminate between novel and familiar conspecifics.

Taken together, vHPC-LS chemogenetic inhibition disrupted the ability of mice to preferentially investigate a novel conspecific over a familiar conspecific while not affecting their ability to discriminate between novel and familiar foods. However, it is unclear whether chemogenetic inhibition increases the preference for a familiar mouse or decreases preference for a novel mouse. To distinguish between these possibilities, we next optogenetically silenced the activity of vHPC-LS neurons in a temporally and spatially restricted manner.

To determine whether inhibition of vHPC-LS neurons differentially influences the investigation of a novel versus a familiar mouse, we used halorhodopsin (NpHR) to silence vHPC-LS neurons in a temporally and spatially specific manner. We injected wild-type mice with a retrogradely transporting Cre virus in the LS and a Cre-dependent NpHR-EYFP or EGFP-only virus in the vHPC and then implanted fibers in the vHPC to target vHPC-LS cell bodies (Figure 2A, Figure 2 - Figure Supplement 1). Using a closed-loop optogenetic manipulation (See Methods), we selectively inhibited vHPC-LS neurons when a conspecific entered a previously defined social zone (Figure 2C).

Interestingly, we found that inhibition of vHPC-LS neurons in the proximity of a familiar mouse abolished the natural preference mice have for novel conspecifics (Figure 2F and G; two-factor ANOVA with light condition and conspecific identity as factors; interaction: p=0.024, main effect of conspecific identity: p=0.006, Sidak post hoc test; OFF: p=0.006, N-ON: p=0.037, F-ON: p=0.949; discrimination score: one sample t-test, OFF: p=0.002, NON: p=0.19, FON: p=0.719). Significantly, light stimulation did not affect the preference for the novel mouse in the control-EGFP group (Figure 2D and E; two-factor ANOVA with light condition and conspecific identity as factors; interaction: p=0.898, main effect of conspecific identity: p=2.8E-05, Sidak post hoc test; OFF: p=0.032, N-ON: p=0.012, F-ON: p=0.032; discrimination score: one sample t-test, OFF: p=0.008, NON: p=0.009, FON: p=0.028). These findings raise the possibility that inhibiting vHPC-LS neurons in the proximity of the familiar mouse could drive increased investigation of that mouse, thus decreasing the time spent in the proximity of a novel mouse.

To determine if silencing of vHPC-LS neurons increased investigation of the mouse paired with stimulation, we inhibited vHPC-LS neurons of test mice in the proximity of one of two novel mice in an arena (Figure 2H–K). We found that NpHR-injected mice spent more time in the proximity of the novel mouse that was paired with inhibition than the other novel animal (Figure 2J and K; two-factor ANOVA with light condition and conspecific identity as factors; interaction: p=0.040, main effect of conspecific identity p=0.005; post hoc Sidak tests; N-OFF: p=0.961, N-ON: p=0.001; discrimination score: one sample t-test, OFF: p=0.796, NON: p=0.029). The control-EGFP group showed an equal preference for both novel animals regardless of the light condition (Figure 2H, I; two-factor ANOVA with light condition and conspecific identity as factors; interaction: p=0.006, no main effect of conspecific identity: p=0.608; discrimination score: one sample t-test, OFF: p=0.535, NON: p=0.128).

Importantly, we found that the increased investigation observed with inhibition of vHPC-LS neurons was socially specific. Inhibition of vHPC-LS activity in the proximity of a novel object did not increase the time spent in the proximity of the object (Figure 2N and O; two-factor ANOVA with light condition and object identity as factors; interaction: p=0.871, no main effect object identity: p=0.505; discrimination score: one sample t-test, OFF: p=0.720, NON: p=0.905). Light stimulation also did not change preferences in the control-EGFP group when presented with two novel objects (Figure 2L and M; two-factor ANOVA with light condition and object identity as factors; interaction: p=0.878, no main effect of object identity: p=0.220; discrimination score: one sample t-test, OFF: p=0.697, N-ON: = 0.422). These findings suggest that the increased investigation of a mouse paired with vHPC-LS inhibition was not simply a result of the inhibition being appetitive.

To determine if changes in locomotion or anxiety-related behaviors could drive increased investigation of the mouse paired with inhibition, we compared both the time spent in the center and velocity of the animals with and without inhibition of the vHPC-LS neurons while mice explored an open field. We found that inhibition of vHPC-LS neurons did not affect either the velocity or the time spent in the center of the arena in the open field assay (Figure 2R and S; one-way ANOVA, time spent: p=0.254; velocity: p=0.775). Light stimulation also had no effect on time spent in the center of the open field arena and the velocity of the animals in the control-EGFP group (Figure 2P and Q; one-way ANOVA, time spent: p=0.094; velocity: p=0.259).

Together, these results suggest that vHPC-LS inhibition likely disrupts preference for a novel conspecific by increasing investigation of a familiar conspecific. Thus, demonstrating that intact activity in the vHPC-LS pathway is necessary for mice to preferentially engage in social novelty-related behaviors. These findings raise the possibility that inhibition of the vHPC-LS neurons could disinhibit downstream regions that have been implicated in promoting approach of novel conspecifics, like the VTA (Smith et al., 2017; Solié et al., 2022; Gunaydin et al., 2014; Bariselli et al., 2018; Bian et al., 2022; Shan et al., 2023), by decreasing the inhibitory drive that GABAergic LS projection neurons exert on their downstream targets.

We hypothesized that effects observed with vHPC-LS inhibition on social novelty-related approach behaviors could be mediated via LS projections to the VTA. To test this hypothesis, we first had to establish the existence of the vHPC-LS-VTA circuitry. To confirm that LS neurons do project to the VTA, we injected a retrogradely transported Cre virus into the VTA and a Cre-dependent GFP virus into the LS (Figure 3A and B). From dense GFP labeling in the LS, we were able to confirm that LS neurons projected to the VTA (Figure 3C).

Next, we determined if LS-VTA neurons received monosynaptic inputs from the vHPC using a modified monosynaptic rabies tracing technique (Wickersham et al., 2007). Three weeks after injecting a retrogradely transporting Cre virus in the VTA and a Cre-dependent rabies helper virus in the LS, we injected the modified delta G-deleted rabies virus into the LS (Figure 3D–F). Using the semi-automated WholeBrain software (Fürth et al., 2018), we mapped brain-wide monosynaptic inputs onto LS-VTA neurons (Figure 3G). We quantified the proportion of total neurons from each region that directly synapse onto LS-VTA neurons. The hippocampal formation was the largest source of inputs to LS-VTA neurons (Figure 3H; one-way ANOVA: p=0.0107, Post hoc multiple comparison Dunnett’s test; HPF vs IC: p=0.1579, HPF vs HY: p=0.0206, HPF vs TH: p=0.0056). Within the hippocampus, the CA1 and CA3 neurons projected most densely onto LS-VTA neurons (Figure 3I; one-way ANOVA: p=0.0096, Post hoc multiple comparison Dunnett’s test; CA1 vs DG: p=0.0411, CA1 vs SUBd: p=0.0511, CA3 vs DG: p=0.0134, CA3 vs SUBd: p=0.0166, CA3 vs SUBv: p=0.0325.). When looking at the organization of inputs along the dorsoventral axis of the hippocampus, we observed that significantly more vCA1 than dCA1 neurons project to LS-VTA neurons (Figure 3J; paired t-test: p=0.0496). In contrast, an equivalent proportion of both dCA3 and vCA3 neurons project onto LS-VTA neurons (Figure 3K; paired t-test: p=0.7736).

From rabies tracing experiments, we have established that the vHPC neurons, specifically neurons from the vCA1 and vCA3 subregions, make dense and direct monosynaptic connections onto LS neurons that in turn project to the VTA. Thus, establishing the existence of a hippocampal-septal-ventral tegmental area circuit that could play a critical role in shaping social novelty discrimination.

After establishing that LS-VTA neurons do receive monosynaptic inputs from the vHPC, we next asked if LS projections to the VTA played a causal role in allowing mice to preferentially investigate a novel conspecific. To test this possibility, we chemogenetically silenced LS-VTA neurons while mice explored the SDT arena containing a novel and familiar conspecific. We injected a retrogradely transporting Cre virus in the VTA of C57BL/6 J mice and a Cre-dependent inhibitory (hM4Di) DREADD virus or an mCherry virus in the LS (Figure 4A and B). Three weeks after surgery, mice received i.p. injections of either saline or CNO to 30 min prior to being run on the SDT (Figure 4C).

Interestingly, inhibition of hM4Di expressing LS-VTA neurons with CNO strongly disrupted the ability of mice to preferentially investigate the novel conspecific in the SDT (Figure 4F and G; two-factor ANOVA with drug condition and conspecific identity as factors; interaction: p=5.1E-04, post hoc Sidak tests: Saline: p=0.001, CNO: p=0.091; discrimination score: one sample t-test, Saline: p=0.012, CNO: p=0.208). Importantly, control (mCherry) mice exhibited a strong preference for the novel conspecific over the familiar conspecific in both saline and CNO conditions (Figure 4D and E; two-factor ANOVA with drug condition and conspecific identity as factors; interaction: p=0.307, main effect of conspecific identity p=6.8E-04, post hoc Sidak tests; Saline: p=2.2E-04, CNO: p=0.0031; discrimination score: one sample t-test, Saline: p=0.045, CNO:p=0.032).

Surprisingly, we found that, unlike with vHPC-LS inhibition, inhibiting LS-VTA neurons appears to cause a trend towards disrupting the mice’s normal preference for familiar foods. Inhibiting the activity of LS-VTA neurons caused mice to trend towards spending equal amounts of time in the proximity of novel and familiar foods (Figure 4J and K; two-factor ANOVA with drug condition and food identity as factors; interaction: p=0.076; discrimination score: one sample t-test, Saline:p=0.065, CNO: p=0.863). Control animals preferred the familiar food in both saline and CNO conditions (Figure 4H, I; two-factor ANOVA with drug condition and food identity as factors; interaction: p=0.941; main effect of food preference: p=2.7 E-05; discrimination score: one sample t-test, Saline: p=0.007, CNO: p=0.055).

Additionally, we were able to confirm that these effects observed with LS-VTA inhibition were not driven by non-specific effects on anxiety-related behaviors. We observed that CNO administration had no effect on the time spent in the center of the open field arena in both control and hM4Di animals (Figure 4L and N; paired t-test, mCherry: p=0.2196, hM4Di: p=0.2182). We also observed that CNO administration did not affect speed of the animals in both control and hM4Di animals (Figure 4M and O; paired t-test, mCherry: p=0.0733, NpHR: p=0.7193).

These findings provide compelling evidence that the increased investigation observed with vHPC-LS inhibition is likely mediated via its projection onto the VTA. Dopamine neurons in the VTA show increased activity while mice investigate a novel conspecific and are known to be involved in causally driving increased investigation of a novel conspecific (Smith et al., 2017; Solié et al., 2022; Gunaydin et al., 2014; Bariselli et al., 2018; Bian et al., 2022; Shan et al., 2023). Together, these findings raise the intriguing possibility that silencing vHPC-LS neurons could disinhibit dopamine neurons in the VTA, which, in turn, could drive increased approach and investigation of a conspecific. Although we have established LS projects to the VTA, it is unclear if LS projects to the dopaminergic neurons in the VTA.

To determine if LS neurons synapse directly onto dopamine neurons in the VTA, we mapped monosynaptic inputs onto dopaminergic neurons in the VTA (VTA_DA) by applying the modified rabies tracing method in Th-Cre⁺ mice. First, we injected a Cre-dependent helper AAV virus into the VTA of Th-Cre⁺ mice. After a 3 week period, we injected a delta G-deleted rabies virus expressing mCherry into the VTA (Figure 5A–C). Using the semi-automated WholeBrain software (Fürth et al., 2018), we mapped and registered mCherry-labeled cells along the entire anterior-posterior axis of the LS (Figure 5D). These mCherry-labeled cells signify LS neurons that make monosynaptic connections onto VTA_DA neurons. We next quantified the anatomical distribution of VTA_DA projecting LS neurons in the brain (Figure 5E–G). Amongst the various subdivisions within the LS, we found that the rostral subdivision of the LS (LSr) projected most strongly to the VTA (Figure 5H; one-way ANOVA: p=5E-06, post hoc Tukey test; LSr vs LSv: p=1.2E-05, LSr vs LSv: p=6E-06, LSc vs LSv: p=0.26). Thus, we have identified a novel pathway that connects LS neurons directly to dopaminergic VTA neurons that could serve as the neural substrate to drive social novelty-related behaviors.

All experimental procedures were conducted in accordance with the National Institutes of Health and Emory University’s Institutional Animal Care and Use Committee (IACUC). All behavioral experiments were performed on male and female C57BL/6 J (Jax strain Number: 000664) mice aged ~4–12 weeks. Rabies tracing experiments included in Figures 3 and 5 were performed on male mice with the animals in Figure 5 belonging to the transgenic strains TH-Cre⁺ (Jax strain Number: 008601). Mice were pair-housed or group-housed in cages with ad libitum access to water and chow (LabDiet). Mice were maintained on a 12 hr reverse light cycle. All experiments were conducted during their light-off period. In social discrimination tests, stimulus mice were sex and age-matched to the test animal.

Mice were placed in a stereotaxic setup (Kopf) and anesthetized with 1–2% isoflurane during surgery. All coordinates were sourced from Paxinos and Franklin, 2019 and are described relative to bregma. Injections were performed using a Nanoject III (Drummond Scientific) and the virus was delivered at a rate of 2 nL per second. Virus injection coordinates are as follows, vHPC: 3 mm posterior, 3.25 mm lateral and –4.2 mm ventral relative to bregma, LS: 0.4 mm anterior, 0 mm lateral, –2.8 mm ventral relative to bregma, VTA: 3.1 mm posterior, 0.35 mm lateral, –4.7 mm ventral.

For vHPC-LS chemogenetic silencing experiments, we bilaterally injected RetroAAV-hSyn-Cre (500 nL, Addgene Lot v70508, 3*10 Besnard and Leroy, 2022) into the LS. This was followed by 250 nl injections of either AAV5-hSyn-DIO-hM4D(Gi)-mCherry or AAV5-hSyn-DIO-mCherry into the vHPC of mice aged 5–6 weeks.

For LS-VTA chemogenetic silencing experiments, we bilaterally injected RetroAAV-hSyn-Cre (500 nL) into the VTA. This was followed by 250 nl injections of either AAV5-hSyn-DIO-mCherry or AAV5-hSyn-DIO-hM4D(Gi)-mCherry into the LS of mice aged 5–6 weeks.

For vHPC-LS cell body inhibition experiments, we bilaterally injected with RetroAAV-hSyn-Cre (500 nL, Addgene Lot v70508, 3*1013) into the LS. This was followed by 500 nL injections of either AAV5-Ef1-DIO-eNpHR3.0-EGFP (Addgene v32533, 1.1*10 13) or AAV5-hSyn-DIO-EGFP (Addgene, 1.1*1013) into the vHPC of mice aged 5–6 weeks. Optical fibers attached to ferrules were secured bilaterally using Metabond and dental acrylic to the skull to target the vHPC (Ferrule coordinates: –3.250 anterior, +/-3.0 lateral, –4.0 depth). Animals were allowed to recover and express the virus for 3 weeks prior to behavioral experiments.

To determine if LS neurons project to the VTA, we injected 750 nl of the retroAAV-Ef1a-mCherry-IRES-Cre virus into the VTA and 750 nl of the AAV5-CAG-FLEX-EGFP into the LS. Animals were allowed to recover and express the virus for 3 weeks prior to histology.

For the LS-VTA monosynaptic rabies tracing experiments, 750 nl of the retroAAV-hsyn-Cre (500 nL, Addgene Lot v70508, 3×1013) was injected in the VTA of male and female C57BL/6 J mice. This was combined with the injection of 750 nl of the helper virus, AAV1.synP.FLEX.splitTVA.EGFP.B19G (Addgene, 2.4×1013) in the LS. Three weeks following injection, all mice were injected with 750 nl of the N2c-ΔG-deleted rabies virus (5×10⁸, Thomas Jefferson University) (Reardon et al., 2016) expressing mCherry and pseudotyped with EnvA, RVdG-mCherry, into the LS.

For VTA_DA monosynaptic rabies tracing experiments, 750 nl of the rabies helper virus, AAV1.synP.FLEX.splitTVA.EGFP.B19G (Addgene, 2.4×1013) was injected into the VTA of Th-Cre⁺ mice (Jax strain number: 008601). Three weeks following injection, all mice were injected with 750 nl of the N2c-ΔG-deleted rabies virus expressing mCherry and pseudotyped with EnvA, RVdG-mCherry into the VTA (~5 × 10⁸, Thomas Jefferson University).

Mice were anesthetized with Euthasol (0.1 mg/kg) then transcardially perfused using PBS (0.5 X) followed by 4% PFA. Brains were extracted, left in PFA overnight, then transferred to a 30% sucrose solution the following day. A microtome (Leica Biosystems) was used to slice the coronal sections (50 um) and validate viral targeting. Slices were mounted using DAPI Fluoromount (Southern Biotech). All mounted sections were imaged using a high-throughput automated stitching fluorescent microscope (Keyence BZ-X810). Native fluorescence of expressed fluorophores was used to confirm expression.

For the chemogenetic silencing experiments, coronal sections of LS (LS-VTA experiments - Figure 4) and coronal sections of vHPC (vHPC-LS experiments - Figure 1) were used to confirm virus expression. Six mice were excluded from the vHPC-LS cohort for showing no expression or off-targeting. Seven mice were excluded from the LS-VTA cohort for showing no expression or off-targeting.

For cell body optogenetic inhibition experiments (Figure 2), coronal sections of vHPC were used to confirm virus expression, nuclear exclusion of NpHR, and to verify targeting of the fibers to vHPC (Figure 2 - Figure Supplement 1). Four mice were excluded from the vHPC-LS optogenetic cohort for showing no expression or off-targeting.

For monosynaptic rabies tracing experiments (Figures 3 and 5), brain sections ranging from the entire brain (~3.5 mm anterior bregma to ~4.5 mm posterior bregma) in ~100 um spacing were mounted. Sections were coverslipped with a DAPI Fluoromount mounting media and imaged using an automated widefield whole slide scanner (Keyence BZ810). Input neurons were mCherry labeled and starter cells expressed both EYFP (helper virus) and mCherry (rabies virus) (Figures 3 and 5).

Arena and handling procedures are identical to SDT. Familiar food was the chow (3 g, LabDiet), while the novel food was 3 g of fruit-flavored cereal. Food was placed under wired cages, such that mice could sniff but not taste the food. Mice were given 5 min to explore the arena containing the novel and familiar food. Percent time spent around the novel and familiar food was calculated by measuring the amount of time the test mouse spent around the cage containing the food (diameter of food zone - 6 cm) relative to the entire assay duration.

Arena and handling procedures are identical to SDT. An object was placed into the home cage of test mice for 72 hr of familiarization. The objects were placed on the opposite ends of the discrimination chamber (without wired cages), and mice were given 5 min to explore the novel and familiar objects. For chemogenetic manipulation, procedures were performed identically to the SDT. Novel and familiar objects were placed on either the left or the right side of the chamber in a counterbalanced fashion.

For spatially-restricted optogenetic manipulation, mice were allowed to investigate two novel objects in an arena for 5 min. In the light ON condition, one of the two objects picked randomly was paired with green light stimulation (532 nm, 6 mW at the fiber tip constant light stimulation). Behavior recording conditions and laser conditions were identical to SDT.

Mice were allowed to freely explore a large square chamber (46.64 cm × 46.64 cm) for 10 min. We determined the amount of time mice spent in the middle of the arena (11.14 cm × 11.14 cm) using Bonsai and MATLAB. For chemogenetic experiments, mice were injected with either saline or CNO 30 min prior to the test. For the vHPC-LS cell body inhibition experiment, the open field task was 9 min long. Mice performed the assay for 3 min with no light stimulation, followed by 3 min of green light stimulation (532 nm, 6 mW at the tip of the optical fiber; constant light illumination) and terminated with an additional 3 min of no light.

The velocity of animals within the Open Field task was determined by calculating distance traveled per unit of time and centroid positioning using Bonsai, an open-source tracking software (Lopes et al., 2015). Velocity was determined by change in pixels moved over change in seconds.

The open-source WholeBrain software package in R allows for the annotation, analysis, and visualization of cellular resolution tracing in an interactive brain atlas (Fürth et al., 2018). The Whole Brain software was used specifically to register individual brain sections from animals onto the Allen common coordinate framework. It was used to map input neurons in rabies tracing experiments (Figures 3 and 5).

In the projection-specific inactivation experiments, all mice received CNO (Sigma; 1 mg/kg, i.p., in 2% DMSO and saline, 1 ml/ 100 g), regardless of virus condition, to equally expose animals to any unintended consequences of CNO. CNO was always administered 30 min prior to the behavioral experiment.

Analyses were performed using PRISM GraphPad, MATLAB, and RStudio. One-way ANOVA or unpaired t-tests were used to compare discrimination scores and velocity. Two-factor ANOVAs were used for social discrimination scores and open-field analyses. In the case of unequal variances, Welch’s ANOVA was used. Tukey’s post hoc tests or paired t-tests were used in the case of significant interactions or main effects with >2 groups and are indicated in the figure with p-value annotations being *p<0.05, **p<0.01, ***p<0.001. Unless otherwise indicated, all tests are two-tailed. Please see Supplementary file 1 for complete statistical analyses.

No new materials were generated during the course of the study.