In initial experiments, we asked whether we could observe changes in the abundance of receptors responsive to an odor after olfactory fear conditioning. The olfactory receptor M71 responds to the odor acetophenone, whereas the receptor MOR23 responds to the odor lyral (Touhara et al., 1999; Bozza et al., 2002). Transgenic mice with targeted mutations at the M71 or MOR23 locus to express green fluorescent protein (GFP) allowed for determination of M71 and MOR23 OSN abundance. Male and female Olfr151^{IRES-tauGFP/IRES-tauGFP} (M71-GFP) or Olfr16^{IRES-tauGFP/IRES-tauGFP} (MOR23-GFP) mice were subjected to an aversive olfactory conditioning paradigm in which a 10 s presentation of the odor acetophenone or lyral was paired with 0.75 mA foot shock, five times daily for 3 consecutive days (Figure 1B and C). The unpaired control group received the same number of odor presentations but experienced a 60-s delay between odor presentation and foot shock (Figure 1C), while the naive control group received no conditioning. The MOEs of conditioned mice were surgically extracted 21 days after the initiation of conditioning and subjected to iDISCO+ optical tissue clearing to visualize M71- and MOR23-expressing OSNs in intact olfactory epithelia (Figure 1B, D and E; Renier et al., 2016). Next, we imaged the anterior region of cleared epithelia using light sheet microscopy and counted the number of M71 or MOR23 OSNs in a fixed volume of tissue using automated spot detection software (Figure 1G and I; Videos 1–6).

Importantly, both M71 and MOR23 OSNs are expressed in the same anterior region of the MOE, enabling consistent imaging and counting protocols for both OSN populations. Male and female M71-GFP mice paired with acetophenone exhibited a 33% increase in the number of M71 OSNs 21 days after the initiation of conditioning when compared to unpaired controls (Figure 1H; One-way ANOVA. p<0.0001. Tukey’s multiple comparisons. Naive vs. F0 paired p<0.0001. F0 unpaired vs. F0 paired p<0.0001. n=12,11,12; Figure 1—figure supplement 1). Male and female MOR23-GFP mice conditioned with lyral exhibited a 39% increase in MOR23 OSNs when compared to unpaired controls (Figure 1J; One-way ANOVA. p<0.0001. Tukey’s multiple comparisons. Naive vs. F0 paired p<0.0001. F0 unpaired vs. F0 paired p<0.0001. n=7,9,9.; Figure 1—figure supplement 1). These results corroborate previous studies and demonstrate that odor-shock pairing leads to an increase in OSN populations that respond to the conditioning odor.

We next asked whether the increase in the number of odor-responsive OSNs following conditioning is inherited by naive offspring of conditioned male mice. Ten days after the initiation of conditioning, we bred F0 males from both the unpaired and paired groups with naive female mice of matching genotypes (M71-GFP or MOR23-GFP). Each mating pair was separated ten days after co-housing to ensure that the offspring were never exposed to the conditioned father. Thus, any observed differences in the offspring could be attributed to the contents of the F0 male’s germline. We collected and optically cleared MOEs from male and female 8- to 10-week-old F1 offspring and compared the abundance of M71 or MOR23 OSNs. These F1 mice were never exposed to acetophenone or lyral, nor had they undergone olfactory fear conditioning. Nonetheless, we observed a 36% increase in M71 OSNs in male and female F1 mice from fathers that underwent paired fear conditioning with acetophenone (Figure 1H; Tukey’s multiple comparisons. F1 unpaired vs. F1 paired p<0.0001. n=12,14.; Figure 1—figure supplements 1 and 2). A similar relative increase of 27% was observed in MOR23 OSNs in male and female F1 mice from fathers that underwent paired fear conditioning with lyral (Figure 1J; Tukey’s multiple comparisons. F1 unpaired vs. F1 paired p=0.0002. n=6,6.; Figure 1—figure supplement 1). These results demonstrate the intergenerational epigenetic inheritance of an olfactory phenotype, namely an increase in specific OSNs in naive F1 offspring following aversive conditioning in F0.

In order to understand how a parental phenotype can be transmitted intergenerationally to future offspring, we decided to first focus on the mechanism underlying the parental phenotype, as it may serve as the basis for the intergenerational mechanism. To investigate the increase in cell number in the parental generation (F0), we first performed a series of control experiments to demonstrate that fear conditioning does not simply lead to a global increase in the number of OSNs in the MOE. First, we took advantage of the fact that the odor propanol does not activate M71 OSNs (Jones et al., 2008; Johnson and Michael, 2000). We repeated the same olfactory conditioning paradigm with homozygous M71-GFP mice, except with propanol as the conditioned odor, and counted the number of M71 OSNs in iDISCO-cleared intact MOEs (Figure 2A). We observed no difference in the number of M71 OSNs between the two groups of mice, indicating that an increase in OSN subtype depends on the conditioned odor activating that OSN population (Figure 2D; Student’s unpaired t-test. Unpaired vs. paired p=0.3009. n=6,7.).

Next, we generated an Olfr151^IRES-tauRFP2 transgenic mouse line that expresses a tau-fused red fluorescent protein (RFP) in OSNs that express the M71 olfactory receptor (Figure 2B). To generate mice with an internal control, we crossed Olfr151^{IRES-tauRFP2/IRES-tauRFP2} (M71-RFP) mice with MOR23-GFP mice to generate an Olfr151^{IRES-tauRFP2/IRES-tauRFP2}; Olfr16^{IRES-tauGFP/IRES-tauGFP} line in which all M71 OSNs are labeled red and all MOR23 OSNs are labeled green (Figure 2C). We conditioned these mice with acetophenone and counted the ratio of M71 OSNs to MOR23 OSNs in serial MOE sections. If olfactory fear conditioning leads to a global increase in the number of OSNs in the MOE, we would expect the ratio of M71 to MOR23 OSNs to remain unchanged following conditioning. However, we observed a significant increase in the ratio of M71 to MOR23 OSNs in paired mice only, demonstrating that odor-shock pairing increased the abundance of M71 OSNs relative to a population that does not respond to the conditioned odor (Figure 2E; One-way ANOVA. p=0.0062. Tukey’s multiple comparisons. Naive vs. paired p=0.0097. Unpaired vs. paired p=0.0163. n=2,4,4.). Taken together, these experiments indicate that olfactory fear conditioning results in a specific increase in the number of cells responsive to the conditioned odor.

The olfactory epithelium exhibits continuous neurogenesis in mice. This constant turnover of OSNs suggests a possible mechanism for the increase in specific OSN populations responsive to the conditioned odor. The increase in M71 and MOR23 cells following odor-shock pairing could result from a biased increase in either the birth or survival rate of specific OSN subtypes. In initial experiments, we examined the relative number of M71 and MOR23 OSNs born during and immediately after olfactory fear conditioning. We injected homozygous M71-GFP and MOR23-GFP mice intraperitoneally with 5-Ethynyl-2′-deoxyuridine (EdU), a thymidine analog that incorporates into newly synthesized DNA and labels newborn cells, during each of the 3 days of training and for 2 subsequent days (Figure 3A). On day 21, we quantified the number of EdU-labeled M71 and MOR23 OSNs in MOE sections (Figure 3A). Since EdU has a half-life of approximately 35 min (Cheraghali et al., 1995), analysis of EdU 16 days after the cessation of injections reflects a pulse-chase, allowing us to quantify a subset of the neurons born during and 2 days following olfactory fear conditioning (Figure 3B).

The number of newborn M71 cells (EdU-labeled) out of total M71 cells was 1.24%±0.29 in naive mice, 2.91%±0.56 in acetophenone-unpaired mice, and 7.61%±0.53 in acetophenone-paired mice (Figure 3D; One-way ANOVA. p<0.0001. Tukey’s multiple comparisons. Naive vs. paired p<0.0001. Unpaired vs. paired p<0.0001. n=6,6,6.). When lyral was used as the conditioned odor, the number of newborn (EdU-labeled) MOR23 cells out of total MOR23 cells was 0.29%±0.06 in naive mice, 0.55%±0.09 in lyral-unpaired mice, and 1.11%±0.21 in lyral-paired mice (Figure 3E; One-way ANOVA. p=0.0120. Tukey’s multiple comparisons. Naive vs. paired p=0.0154. Unpaired vs. paired p=0.0653. n=4,6,8.). For both conditioning odors, we observed a significant increase in the number of newborn odor-responsive OSNs in mice that underwent paired conditioning (Figure 3D and E). These observations demonstrate that olfactory fear conditioning with acetophenone and lyral results in a significant increase in the number of newborn M71 and MOR23 cells, respectively. Despite subtype-specific differences in proliferation rates, odor-shock pairing significantly increased the number of newborn cells differentiating into an OSN subtype that responds to the conditioned odor.

We observe an increase in M71 OSNs 21 days after the start of acetophenone-paired olfactory fear conditioning. Given the continuous regeneration of the MOE, we next asked if this increase persists at later time points (Figure 4A). Since the half-life of the mouse MOE (the amount of time required for half of the epithelium to regenerate) is approximately 26 days (Holl, 2018), then at 42 days (6 weeks), approximately 67% will have been replaced by newly born neurons, and at 63 days (9 weeks), approximately 81% will have been replaced. At day 42, 3 weeks after the time point at which M71 counts were performed in initial experiments, we extracted the MOEs of male and female homozygous M71-GFP mice that underwent either unpaired or paired conditioning paradigms and optically cleared the tissue to compare M71 OSN abundance (Figure 4C). At the 42-day time point, we observed a 20% increase in the abundance of M71 OSNs in paired animals compared to controls (Figure 4C; Tukey’s multiple comparisons. 42d unpaired vs. paired p=0.0476. n=8,8. Figure 4—figure supplement 1). At 63 days, we observed a 30% increase in M71 OSNs in the paired group compared to controls (Figure 4C; Tukey’s multiple comparisons. 63d unpaired vs. paired p=0.0011. n=4,6.; Figure 4—figure supplement 1). The observation that these changes persist for at least 63 days after the start of conditioning, together with the reported 26 day half-life for the MOE, suggests the persistence of a signaling mechanism past the experience of olfactory fear conditioning itself.

To test whether a behavioral phenotype depends on OSN abundance, we performed an avoidance assay at the 42- and 63-day timepoints in a new cohort of mice. Mice were allowed 10 min to freely explore a three-chamber arena (trichamber) in which the conditioned odor (acetophenone) was continuously delivered through a port on one side, and a control odor (propanol) on the other (Figure 4D; Video 7). We calculated an approach-avoid index as the difference in the amount of time spent on the conditioned odor side compared to the control odor side, normalized by the total time spent exploring either chamber (Figure 4D). A positive index indicates overall approach towards the conditioned odor, whereas a negative index indicates an aversion to the conditioned odor. At the 42-day time point, we observed no significant difference in the avoidance indices of unpaired and paired mice, with both groups exhibiting a slight aversion to acetophenone (Figure 4E; Student’s unpaired t-test. 42d unpaired vs. paired p=0.2248. n=10,11.). At the 63-day time point, we also observed no significant difference between the unpaired and paired mice (Figure 4F; Student’s unpaired t-test. 63d unpaired vs paired p=0.8987. n=6,8.). Interestingly, when accounting for sex, we found a significant aversion in female paired mice at the 42-day timepoint that was not present at the 63-day timepoint and not present at either timepoint in males (Figure 4—figure supplement 2). Despite the behavioral sex difference at day 42, we did not observe a significant sex difference in M71 OSN counts at the same timepoint (Figure 4—figure supplement 1), indicating that the difference in behavior was not driven by a difference in MOE composition. Moreover, the absence of a strong behavioral aversion to acetophenone at 6- and 9 weeks post-conditioning despite a persistent increase in M71 representation in the MOE further suggests that an increase in OSN subtype following conditioning does not directly drive avoidance behavior. If this were the case, we would expect that the paired groups, which both exhibit heightened M71 counts at 42 and 63 days, would exhibit greater aversion to acetophenone than unpaired mice. Altogether, these results indicate that the mechanism underlying cellular adaptations in the MOE and the mechanism driving avoidance are independent.

To further disentangle the relationship between increased OSN subtype representation and avoidance behavior, we performed extensive analyses on both F0 and F1 behavior. We generated three parallel F0 and F1 cohorts with three odors: acetophenone, lyral, and propanol. In each cohort of mice in which F0 males were conditioned for F1 breeding, we also conditioned male and female mice and tested them in the trichamber assay the day after conditioning (Figure 5B). We used ANY-maze tracking software to track the center position of the mice for calculation of approach-avoid indices, as well as to measure metrics like speed, distance traveled, and freezing. Importantly, males that were used for F1 breeding were not tested in the trichamber prior to breeding to avoid extinction, which has been shown to reverse the increased OSN phenotype in both F0 and F1 (Aoued et al., 2019).

As expected, only mice in which odor and shock were paired exhibited robust aversion to the conditioned odor. F0 mice in the paired groups actively avoided the conditioned odor, whereas mice in the unpaired group exhibited no aversion to the conditioned odors (Figure 5C, G and K; Tukey’s multiple comparisons. F0 acetophenone unpaired vs. paired p<0.0001. n=10,15. F0 lyral unpaired vs. paired p<0.0001. n=17,20. F0 propanol unpaired vs. paired p<0.0001. n=16,20.). Naive and unpaired mice spent roughly equal time exploring the control and conditioned odor chambers, whereas the paired mice spent an average of 67% (lyral-paired), 75% (acetophenone-paired), and 92% (propanol-paired) of the time exploring the control chamber.

Previous behavioral studies demonstrated that F1 offspring from fathers that experienced olfactory fear conditioning exhibit enhanced sensitivity to the conditioned odor in both odor potentiated startle and aversive odor association assays (Dias and Ressler, 2014). Therefore, we asked whether we could detect aversive behavioral responses in F1 populations after conditioning F0 fathers. We decided to use the trichamber assay as opposed to assays used in previous studies because we were interested in whether offspring inherited, in addition to an increase in conditioned odor OSN representation, a behavioral aversion to the odor. F1 offspring generated from conditioned males and naive females were assayed at 8- to 10 weeks of age in the same trichamber assay as the F0s. Importantly, none of the offspring had ever been exposed to the odor with which their father had been conditioned. We did not observe a significant difference in the avoidance indices of unpaired and paired F1 mice for any of the odors (Figure 5C, G and K; Tukey’s multiple comparisons. Acetophenone F1 unpaired vs. paired p=0.9999. n=22,18. Lyral F1 unpaired vs. paired p=0.9976. n=6,13. Propanol F1 unpaired vs. paired p=0.7935. n=13,13. Figure 5—figure supplements 1–3).

Suspecting that our approach-avoid index may not be sensitive to more subtle behavioral differences, we performed additional behavioral analyses. Generation of spatial heat maps for the assays failed to reveal any clear qualitative differences in occupancy between unpaired and paired F1s in any of the three cohorts (Figure 5D, H and L). We also quantified freezing behavior during the trials and found no differences between unpaired and paired F1s in any of the cohorts (Figure 5F, J and N). However, we identified nuanced differences between unpaired and paired F1s that depended on the father’s conditioned odor. When we quantified the distance traveled during the assay, we found that the F1s of lyral-paired fathers exhibited a hyperactive phenotype, traveling an average of 24.74±0.74 meters during the 10-min trial compared to 19.12±1.66 m in F1 lyral-unpaired mice (Figure 5I; S5E. Tukey’s multiple comparisons. Lyral F1 unpaired vs. paired p=0.0325. n=6,13.). Conversely, the F1s of propanol-paired fathers exhibited a hypoactive phenotype, traveling an average of 15.89±1.11 m compared to 23.33±1.95 m in F1 propanol-unpaired mice (Figure 5M; Tukey’s multiple comparisons. Propanol F1 unpaired vs. paired p=0.0045. n=13,13.; Figure 5—figure supplement 1). The phenotypes observed in the F1s of both lyral-paired and propanol-paired fathers are similarly reflected in the mean speeds of the mice during the trial, both of which were significantly different between the F1 unpaired and paired groups (Figure 5—figure supplement 3; Tukey’s multiple comparisons. Lyral F1 unpaired vs. paired p=0.0315. n=6,13. Propanol F1 unpaired vs. paired p=0.0051. n=13,13.). We observed no significant differences in either distance traveled or mean speed in the F1s of acetophenone-conditioned fathers (Figure 5E; Figure 5—figure supplement 3). These analyses demonstrate that while the offspring of paired fathers do not actively avoid their father’s conditioned odor, odor-shock pairing leads to subtle behavioral differences in offspring, and that these differences are specific to the conditioned odor used.

To follow up on the differences observed between the F1s of unpaired and paired mice, we utilized Keypoint-MoSeq, a tool that uses unsupervised machine learning to identify behavioral modules, called ‘syllables’, in an unbiased manner (Weinreb et al., 2024). First, we trained a SLEAP model to track eight keypoints (nose, right ear, left ear, neck, center, right leg, left leg, tail-base) on each mouse across the entire trial (Figure 6B; Pereira et al., 2022). We trained a Keypoint-MoSeq model on the tracking data from 99 10-min trichamber videos (30 F0 and 69 F1 videos), for a total of 1.78 × 10⁶ frames, to identify behavioral syllables with a median duration of approximately 30 frames (1 s; Figure 6B and C). This allowed us to ask whether the differences we observed between lyral F1s and propanol F1s at the timescale of 10 min could be tied to specific behaviors at the timescale of seconds, rather than a summary of 10 min. In our first analysis, we grouped the F1s of acetophenone-, lyral-, and propanol-unpaired fathers, and grouped the F1s of acetophenone-, lyral-, and propanol-paired fathers, to investigate whether there are differences between unpaired and paired offspring that generalize across the conditioned odors. Strikingly, we identified four syllables (3, 5, 11, and 22) with a significant difference in relative frequency (usage) between unpaired and paired F1s (Figure 6D; Kruskal-Wallis test with Dunn’s multiple comparisons. F1 unpaired vs. paired. Syllable 3 p=0.0070. Syllable 5 p=0.0054. Syllable 11 p=0.0251. Syllable 22 p=0.0184.). Of those four syllables, syllables 3 and 5 were used significantly more in the F1 paired group compared to the F1 unpaired group, while syllables 11 and 22 were used significantly less in the F1 paired group (Figure 6D). We also noticed interesting qualitative differences in the spatial usage patterns of each syllable, with unpaired and paired F1s showing distinct usage distributions relative to their distance from the conditioned odor port (Figure 6—figure supplement 1). Next, we were interested in behavioral differences specifically in response to the conditioned odor of the parent. To this end, we narrowed the analysis to syllable usage only when the mouse was in the conditioned odor chamber. While both F1 unpaired and F1 paired mice spent similar amounts of time in the conditioned odor chamber (Figure 6E; Figure 5—figure supplement 3), the F1 paired group exhibited greater usage of syllables 5 and 17 compared to controls (Figure 6F; Kruskal-Wallis test with Dunn’s multiple comparisons. F1 unpaired vs. paired. Syllable 5 p=0.0406. Syllable 17 p=0.0467.). Taken together, these findings point to small but significant behavioral differences between the offspring of unpaired and paired fathers that generalize across all three conditioning odors.

Lastly, we performed the same analysis separately for the three F1 cohorts of acetophenone-, lyral-, and propanol-conditioned F0 fathers. Interestingly, in parallel with the earlier analyses, we again observed no differences between the unpaired and paired F1s of acetophenone-conditioned fathers (Figure 6G) despite differences in M71 OSN representation. Also consistent with our earlier analyses, we found significant differences between the unpaired and paired F1s of lyral- and propanol-conditioned fathers, and the differences between the unpaired and paired F1 groups were unique to the odor (Figure 6H and I). We identified six syllables with significantly different frequencies between unpaired and paired F1s of lyral-conditioned fathers: 2, 3, 5, 7, 15, and 20 (Figure 6H; Kruskal-Wallis test with Dunn’s multiple comparisons. Lyral F1 unpaired vs. paired. Syllable 2 p=0.0016. Syllable 3 p=0.0272. Syllable 5 p=0.0030. Syllable 7 p=0.0025. Syllable 15 p=0.0208. Syllable 20 p=0.0002.). The upregulated syllables in lyral paired F1s were higher-velocity syllables, while the downregulated syllables were relatively stationary/low-velocity, consistent with the hyperactive phenotype earlier described by greater distance traveled and higher mean speed in lyral-paired F1s. Between the unpaired and paired F1s of propanol-conditioned fathers, we found five syllables with significantly different frequencies: 6, 12, 14, 15, and 20 (Figure 6I; Kruskal-Wallis test with Dunn’s multiple comparisons. Propanol F1 unpaired vs. paired. Syllable 6 p=0.0206. Syllable 12 p=0.0407. Syllable 14 p=0.0072. Syllable 15 p=0.0320. Syllable 20 p=0.0148.). Compared to the lyral F1s, the opposite was true for the differentially used syllables in the propanol F1s, in which the upregulated syllables in propanol-paired F1s were lower-velocity syllables, and the downregulated syllables were relatively high-velocity syllables, consistent with the hypoactive phenotype earlier described. Here, we have demonstrated that olfactory fear conditioning in the F0 population leads to conditioned odor-specific effects in the F1 offspring of paired mice.

All procedures were approved by the Columbia University Institutional Animal Care and Use Committee under protocol #AABL8552. All mice were housed with a 12 hr light/12 hr dark cycle and fed ad libitum. Olfr151^{IRES-tauGFP/IRES-tauGFP} (Stock #006676), Olfr16^{IRES-tauGFP/IRES-tauGFP} (Stock #006643), and C57BL/6 J (Stock #000664) mice were obtained from The Jackson Laboratory or gifted from the Lomvardas and the Axel laboratories. The Olfr151^IRES-tauRFP2 line was generated in-house.

The Olfr151^IRES-tauRFP2 mouse line was generated by homologous recombination of mouse ES cells using established techniques. Briefly, the targeting vector (Addgene, 15510) was linearized using PmeI and electroporated into MM13 ES cells (129 S/SvEv). Targeted ES clones were identified by Southern blot hybridization and positive ES cells were transferred into C57BL/6 blastocysts. Male chimeras were bred with C57BL/6 females to establish germline transmission and subsequently outcrossed to C57BL/6 females for at least five generations. Deletion of the self-excising Neo cassette was confirmed by PCR of genomic DNA.

8–12 week-old male and female mice were trained to associate acetophenone (Sigma-Aldrich, 42163), lyral (IFF, 00129214), or propanol (Sigma-Aldrich, I9516) with 0.75 mA foot shocks. Odors were diluted to 10% v/v in mineral oil (Fisher, O121-1). The mice were trained on 3 consecutive days, with each training day consisting of five presentations of odor for 10 s. For mice in the paired condition, the odor presentations were co-terminated with a 0.75 mA foot shock. For mice in the unpaired condition, there was a 60-s delay between the odor presentation and foot shock (Figure 1C). Olfactory fear conditioning boxes, olfactometers, and software were obtained from Med Associates. Mice were randomly assigned to experimental conditions.

MOEs were perfused, dissected, and processed according to the iDISCO+ protocol (Renier et al., 2016). Whole MOEs were processed in 5 mL volumes. Samples were postfixed in 4% paraformaldehyde in 1X PBS (Electron Microscopy Sciences, 15,710 S) overnight at 4°C. The following day, they were washed with 1X PBS (3 × 30 min), gradually dehydrated with methanol (MeOH; Sigma-Aldrich, 322415) over 5 hr, and incubated in 66% dichloromethane (DCM; Sigma-Aldrich, 270997)/33% MeOH overnight. The samples were washed in 100% MeOH the following day, chilled at 4°C, bleached in 5% hydrogen peroxide (Sigma-Aldrich, 216763) in MeOH overnight at 4°C, and then gradually rehydrated the next day. Samples were permeabilized for 2 days and blocked in 6% goat serum (Jackson ImmunoResearch, 005-000-121) for 2 days at 37°C. Next, they were labeled with a 1:2000 dilution of primary chicken anti-GFP antibody (Aves Labs, GFP-1020, RRID:AB_10000240) for 3 days at 37°C, washed for 1 day (5 ×1 hr), and labeled with a 1:1000 dilution of secondary goat anti-chicken Alexa Fluor 647 antibody (Thermo Fisher Scientific, A-21449, RRID:AB_2535866) for 3 days at 37°C. Due to the fragility of the nasal turbinates housing the majority of zone 1, samples were embedded in 4% agarose (Invitrogen, 16500100) prior to final dehydration (Figure 1E). Lastly, the embedded samples were incubated in 66% DCM/33% MeOH for 3 hr, rinsed twice with 100% DCM, and transferred to dibenzyl ether (Sigma-Aldrich, 33630) for final clearing. This experiment was replicated at least five times across multiple cohorts, with each cohort representing all groups.

Images were collected with a light-sheet microscope (Ultramicroscope II, LaVision BioTec) at ×2.0 magnification using a 640 nm laser and a z-step size of 2.0 µm. All cleared tissue images were acquired with tissue submerged in dibenzyl ether (DBE). Laser intensity was set between 55% and 75% to prevent oversaturated pixels and photobleaching. The working distance of the microscope allowed for complete visualization of both left and right turbinates containing zone 1 OSNs (Figure 1F and H). The light-sheet microscope was provided by Cellular Imaging at the Zuckerman Institute (NIH 1S10OD023587-01).

All quantification was performed in a double-blind manner. The image stack of GFP+ olfactory sensory neurons in the 647 nm channel was analyzed using Imaris software. The average number of olfactory sensory neurons was measured using the spot detection tool on 350³ µm³ cubes of zone 1 tissue, with a requirement of at least 3 cubes per sample for inclusion. The decision to measure an average number of cells within a fixed volume, as opposed to the total number of cells in the turbinates, accounted for potential tissue loss and differences in tissue volumes/shapes across samples. The spot detection was set to a 16.3 µm estimated diameter and was based on Imaris’ quality threshold, which compares the intensities at the centers of the candidate spots. The quality threshold varied slightly across samples to adjust for signal quality and axon brightness (to minimize counting spots on axons) but was held consistent within every sample.

10 mM EdU (Invitrogen, E10187) was administered to male and female 8–12 week-old M71-GFP^+/+ and MOR23-GFP^+/+ mice through a series of daily 0.01 mL/g intraperitoneal injections. Injections were administered 15 min prior to olfactory fear conditioning on each of the 3 days of conditioning, plus 2 additional days around the same time conditioning had been performed (Figure 3A). This experiment was replicated at least three times across multiple cohorts, with each cohort representing all groups.

Mice were transcardially perfused with ice-cold 4% PFA (Electron Microscopy Sciences, 15,710 S) in 1X PBS. MOEs were surgically dissected, incubated in 4% PFA overnight, and cryoprotected in 30% sucrose (Sigma-Aldrich, S0389). MOEs were frozen in OCT (Fisher, 23-730-571) and stored at –20°C until sectioning. Tissue was sliced into 20 µm sections, mounted directly onto Fisher Superfrost Plus glass slides (Fisher, 12-550-15), and stored at –80°C until staining. At the time of staining, slides were acclimated to room temperature, washed with PBST (0.1% Triton X-100 in 1X PBS; 3 × 5 min; Sigma-Aldrich, X100), and incubated with Click-iT Plus EdU reaction cocktail (Alexa Fluor 555; 30 minutes; Invitrogen, C10638). Sections were washed again with PBST (3 × 5 min), with the last wash including 1:10,000 DAPI (Invitrogen, D1306), and then cover-slipped using Vectashield Plus mounting medium (Vector Labs H-1900) and sealed with nail polish.

Slides were imaged using a Zeiss Upright LSM 880 Confocal microscope and Zen Black software (Zeiss) or W1-Yokogawa Spinning Disk Confocal (Nikon). All co-labeling images were acquired in z-stacks to ensure accuracy in co-labeling determination.

At least 1 day following olfactory fear conditioning, conditioned mice were assayed in a custom-built acrylic three-chamber box to assess odor avoidance behavior (Figure 4). The conditioned and control odors were assigned randomly to either side. Odors were diluted to 1% v/v in mineral oil for all odor preference assays, and flowmeters were set to equal flow rates of approximately 1 l/min. The three-chamber box included vacuum ports both in the center chamber, as well as on either side of the center chamber doorways in the side chambers to ensure that each experimental odor was restricted to its chamber. The mice were habituated to the center chamber for 1 min prior to the start of the test, and then the doors to both chambers were lifted to initiate the assay. Mice were recorded roaming freely throughout the three-chamber box for 10 min. This experiment was replicated at least 5 times across multiple cohorts, with each cohort representing all groups.

We used ANY-maze software to track the position of the mice throughout the entire trichamber trial. The odor avoidance index was calculated as approach-avoidance index = (time spent on control odor side - time spent on conditioned odor side) / total time spent on either side. Animals visibly sampled each chamber to be included in analysis. We also used ANY-maze to generate heat maps and determine distance traveled, time freezing, and mean speed metrics during the assay. For pose estimation, we trained a SLEAP model (Pereira et al., 2022) to track eight keypoints (nose, right ear, left ear, neck, center, right leg, left leg, tailbase) on each mouse across each 10-min trial. With tracking data from 99 (30 F0, 69 F1) 10-min trichamber videos that were all recorded by the same camera and in the same apparatus, we trained a Keypoint-MoSeq model (Weinreb et al., 2024) to identify behavioral syllables. The model had six latent dimensions (explaining >90% of the variance) and a kappa value of 1e6. For statistical analyses, a minimum relative frequency cutoff of 0.005 was applied for syllables. Pre-established exclusion criteria required for each mouse to sample each side (control odor and conditioned odor) for at least 3 s for inclusion.

For all F0 experiments, group assignment (naive, unpaired, paired) was randomized. For F1 experiments, group assignment was determined based on the condition of the father used for mating. Power analyses were performed for each experiment using G*Power (3.1) to determine sample sizes. A significance threshold of p<0.05 was used for all statistical analyses. Outlier tests were performed on all data, leading to the removal of one statistical outlier from the Acetophenone F1 Paired group. This outlier was also excluded from analyses of other metrics from the same behavioral assay. After performing Shapiro-Wilk normality tests, normally distributed data across 2+ groups were performed using a standard one-way ANOVA with Tukey’s multiple comparisons post-hoc tests. Normally distributed data between two groups were performed using Student’s unpaired t-tests. For data that was not normally distributed, 2+ groups were analyzed with a nonparametric one-way ANOVA (Kruskal-Wallis test) and Dunn’s multiple comparisons post-hoc tests. Descriptive statistics used standard error of the mean (S.E.M.) to estimate error. Percent differences between two groups were calculated by comparing the mean of each group. For all analysis except Keypoint-MoSeq, statistical analysis was performed using Prism 9 (GraphPad) software. Statistics for all figures and figure supplements can be found in Supplementary files 1 and 2, respectively. For Keypoint-Moseq statistics, a combination of Keypoint-MoSeq code and custom Python code (available on Github at https://github.com/BJMarlinLab/Liff_et_al_2026 copy archived at Marlin Lab, 2026) was used for statistical analysis.

Code is available at https://github.com/BJMarlinLab/Liff_et_al_2026 (copy archived at Marlin Lab, 2026).