Opposing spatial gradients of inhibition and neural activity in mouse olfactory cortex

The spatial representation of stimuli in primary sensory cortices is a convenient scaffold for elucidating the circuit mechanisms underlying sensory processing. In contrast, the anterior piriform cortex (APC) lacks topology for odor identity and appears homogenous in terms of afferent and intracortical excitatory circuitry. Here, we show that an increasing rostral-caudal (RC) gradient of inhibition onto pyramidal cells is commensurate with a decrease in active neurons along the RC axis following exploration of a novel odor environment. This inhibitory gradient is supported by somatostatin interneurons that provide an opposing, rostrally-biased, gradient of inhibition to interneurons. Optogenetic or chemogenetic modulation of somatostatin cells neutralizes the inhibitory gradient onto pyramidal cells. This suggests a novel circuit mechanism whereby opposing spatial gradients of inhibition and disinhibition regulate neural activity along the RC-axis. These findings challenge our current understanding of the spatial profiles of neural circuits and odor processing within APC.


How do inhibitory circuits implement opposing RC inhibitory asymmetries and what is the
PV cell density appears to decrease along the RC axis. However, this was only significant in one PV 149 animal (filled black circle, Figure 3C4, Table 1) and the average normalized density across all animals 150 (-0.27 ± 0.10, p: 0.031, Figure 3C3). In the majority of mice (n=5/6), PV cell density did not significantly 151 vary with RC distance and the distribution of slopes did not differ from zero (-0.16 ± 0.08, p: 0.065,  test, open circles, Figure 3C4). Altogether, these findings suggest it is unlikely that stronger caudal 153 inhibition of dPCs arises from a rostral-to-caudal increase in interneuron density in L3. 154 We have previously shown that SST-cells inhibit the majority of L3 interneurons in APC 42 . 155 This suggests that the high rostral density of SST cells could underlie rostrally-biased inhibition of 156 interneurons. To test this, we selectively expressed ChR2 in SST-cells and repeated grid stimulation 157 while recording IPSCs in L3 interneurons and dPCs ( Figure 4A). As predicted, SST-cells provided 158 rostrally-biased inhibition to the majority (73%) of L3 interneurons (bias: -0.11 ± 0.04, p: 0.02, n=22, 159 Figure 4B1,2). Bias values did not significantly differ between VGAT-ChR2 and SST-ChR2 animals (p: 160 0.69, unpaired t-test). Thus, activating solely SST-cells replicates the rostrally-biased inhibition (I R : 0.58 161 ± 0.03 > I C : 0.048 ± 0.03, p: 0.017, paired t-test, Figure 4D) of L3 interneurons seen in VGAT-ChR2 162 animals. Interestingly, the strength of SST-mediated inhibition onto L3 interneurons also did not 163 significantly differ (6.82 ± 1.4 pAs) from that in VGAT-ChR2 animals (7.02 ± 0.92 pAs, p: 0.91, unpaired 164 t-test). Although one should be cautious comparing between transgenic lines, one interpretation is that 165 SST cells are a major source of rostrally-biased inhibition onto L3 interneurons. 166 In dPCs, SST-ChR2 mediated inhibition was significantly weaker (4.77 ± 0.89 pAs) than 167 VGAT-ChR2 (8.63 ± 0.84 pAs, p: 0.003, unpaired t-test). This would be expected if SST-cells are only a 168 subset of the interneurons that inhibit dPCs. SST-ChR2 cells provided rostrally-biased inhibition to 169 50% of dPCs (n=7/14, Figure 4C1) compared to only 2/16 in VGAT-ChR2 animals. Consequently, the 170 distribution of bias values was not coherently asymmetric and did not differ from zero (0.06 ± 0.08, p: 171 0.45, one sample t-test, n=14, Figure 4C2). Finally, in contrast to VGAT-ChR2 activation, SST-ChR2 172 inhibition did not differ between rostral and caudal sites (p: 0.50 paired t-test, Figure 4D). Thus, 173 activation of solely SST-cells increases rostrally-biased inhibition of many dPCs and ultimately 174 neutralizes inhibitory bias across the population. These findings are consistent with rostrally-biased 175 distributions of SST-cells and suggest that additional inhibitory circuits are required to produce 176 consistent, caudally biased inhibition of dPCs. 177 There are two ways to produce caudally biased inhibition of dPCs -1) increase caudal 178 inhibition; or 2) decrease rostral inhibition. We have not found a mechanism to support increased 179 caudal inhibition. However, we have shown that SST-cells provide rostral inhibition to interneurons 180 ( Figure 4B), which could decrease rostral inhibition of dPCs. To test this possibility, we bred triple 181 transgenic animals that express ChR2 in all interneurons but only SST-cells express the inhibitory 182 DREADD, hM4Di (abbreviated: VGAT-ChR2-SST-Di). In these mice, the DREADD agonist CNO (20 183 µM, bath) reduces SST-cell activity. We performed grid stimulation of L3 sites ( Figure 4E) and 184 compared IPSC strength and RC bias in control conditions (green) versus CNO (black, Figure 4F). 185 Upon application of CNO, IPSC strength decreased consistent with a loss of direct, SST mediated 186 inhibition of dPCs (Control: 4.14 ± 0.70 pAs; CNO: 2.72 ± 0.40 pAs, p: 0.007, paired t-test, n=12). If 187 SST-cells influence caudal bias through rostral disinhibition, we expect a loss in SST-mediated 188 inhibition would shift bias to less caudal values. Surprisingly, the mean bias did not change significantly 189 in CNO (bias: +0.17 ± 0.08, n=12) compared to control (bias: +0.19 ± 0.06, 0.57, paired t-test). 190 However, the bias distribution was significantly asymmetric in control conditions (mean>0, p: 0.010, one 191 sample t-test), whereas in CNO, bias values were not significantly asymmetric (p: 0.051, one sample t-192 test). This is because caudal bias both increased (n=4/12) and decreased (n=8/12) across the dPC 193 population with CNO application. In a small number of dPCs, increased caudal bias (n=4/12, Δ Bias = 194 Bias CNO -Bias Control =+0.13 ± 0.04, gray bars Figure 4G) can be explained by a loss of direct,  mediated inhibition at rostral sites. In contrast, the majority of dPCs (n=8/12) shifted toward less caudal 196 bias values with CNO (Δ Bias -0.09 ± 0.03, black bars, Figure 4G). In these cells, CNO produced a 197 significantly greater reduction in inhibition at caudal sites (-35 ± 9%) versus rostral sites (-24 ± 7%, 198 p<0.05, WSR-test, n=8). This suggests that rostral interneurons are normally suppressed by  in control conditions, but rebound during CNO application and neutralize bias. Thus, caudally biased 200 inhibition of dPCs could arise by rostral disinhibition of PCs through SST-to-interneuron microcircuits.   (Figure 5B1,2). Briefly, HC (n=6) and HCO 212 (n=6) mice were given a single dose of 4-OHT and then returned to the home cage. HCO mice were 213 allowed to rest for 30 min then exposed to odor in the home cage for 30 min. NEO mice (n=6) were 214 given 4-OHT, rested for 30 min, and then explored a novel environment plus odor for 30 min before 215 being returned to the home cage. The novel environment was a divided arena with two cups of 216 bedding-one odorized, one blank-at the end of each arm ( Figure 5B2, schematic far right). In a 217 subset of mice (n=4), location within the arena was monitored. NEO mice were highly active and 218 sampled both arms as well as the center (C) of the arena throughout exposure period (30 min, Figure  219 5B3). Mice spent nearly equivalent time per visit in the odorized (8.1 ± 3.5 s) and blank (9.8 ± 3.5 s) 220 arms, but visited the blank arm (49 ± 33 visits) more frequently than the odorized arm (25 ± 14 visits). 221 The majority of HCO and NEO mice were exposed to isoamyl acetate. One cohort of mice (n=2 HCO, 222 n=1 NEO) was exposed to ethyl-butyrate. Results did not differ between odors and were grouped. 223 Following odor exposure, mice remained in their home cages, undisturbed, in the dark for 10-12 hours. 224 Mice were sacrificed 5 days later and tdTom(+) neurons were counted along ~1.5 mm of the RC axis. 225 Neural activity was quantified as the density (cells/mm 2 ) of tdTom(+) cells in laminar regions of interest 226 (L2, L3 ROIs, Figure 5A) located directly under the LOT. Densities were normalized to the most rostral 227 section for linear fits as described for interneuron densities (Figure 3). Summary data is presented in 228 We found laminar differences in both the average density and the RC spatial pattern of active 231 neurons with odor exposure. In L2, the average density of tdTom(+) neurons (in cells/mm 2 ) was 232 significantly greater in odor-exposed animals, NEO (243 ± 43) and HCO (199 ± 13), compared to HC 233 animals (113 ± 18, p: 0.013 KW-test, Figure 5F1). In L3, average density did not vary with condition 234 (HC: 52 ± 11; HCO: 68 ± 5, and NEO: 76 ± 15, p: 0.149 KW-test, Figure 5F2). In contrast, we found 235 RC spatial patterning of neural activity in L3 ( Figure 5D) but not L2 ( Figure 5C). The normalized 236 density of tdTom(+) neurons was plotted against RC distance for individual mice (open triangles) and 237 averaged across animals (solid triangles, Figure 5C,D). RC patterning was defined as significant non-238 zero slope values from linear fits of RC densities in individual animals as well as across animals within 239 a group (Table 2). In L2, the distributions of slope values did not significantly differ from zero (HC: 0.39 240 ± 0.27; HCO: -0.01 ± 0.10; NEO: -0.08 ± 0.07 mm -1 , p: 0.065 -0.38, MWU-test, Figure 5C, Table 2) or 241 between conditions (p: 0.104 KW-test, Figure 5E1). However, in L3, there was significant RC 242 patterning of active neurons that further differed between HC, HCO and NEO conditions ( Figure 5D, 243 E2). All NEO mice showed a significant decrease in the density of active neurons along the RC axis 244 (filled red triangles, slope distribution ≠ 0: -0.45 ± 0.05 mm -1 , **p: 0.005, MWU-test, Figure 5E2, Table  245 2). Further, the distribution of slope values was significantly more negative in NEO mice than HCO or 246 HC (★★, p: 0.0046, KW-test, Figure 5E2). Consistent with findings from individual NEO animals, the 247 average change in RC density across animals was also significantly negative (red triangles, -0.47 ± 248 0.04, p<0.000, Figure 5D3). In HCO animals, RC patterning was shallower and less reliable in 249 individual mice than NEO animals. The average density across animals decreased significantly with RC 250 distance but the slope was less than half that of NEO animals (gold triangles: -0.19 ± 0.04; p: 0.0005, 251 linear regression, Figure 5D2). Further, although the distribution of slopes across mice was significantly 252 non-zero (-0.21 ± 0.12 mm -1 , **p: 0.005, MWU-test), RC decreases were rarely significant in individual 253 mice (n=2, filled orange triangles, Figure 5E2, Table 2). In HC animals, the average density across 254 animals appears to increase with RC distance (slope: 0.24 ± 0.10 mm -1 ; p: 0.037, linear regression, 255 Figure 5D1). However, in individual animals, the distribution of slopes was inconsistent-positive (n=3), 256 negative (n=2) and neutral (n=1) ( Table 2) and did not significantly differ from zero (slope: 0.13 ± 0.14 257 mm -1 , Figure 5E2). Finally, across conditions, changes tdTom(+) densities do not correlate with 258 changes in the total number of cells along the RC axis. In 6 mice, two from each group (NEO, HCO, 259 HC) we quantified the RC density of all cells (DAPI stain) in L2 and L3 (Supplemental Figure 2). Total 260 cells consistently increased along the RC axis in L2 (slope distribution ≠ 0: 0.54 ± 0.21 mm -1 , p: 0.0051 261 MWU) but not in L3 (0.08 ± 0.11 mm -1 , p: 0.94 MWU, n=6). To summarize, we show that odor exposure 262 increases the density of active neurons in L2 but not L3, and significantly changes the RC spatial 263 patterning of neural activity in L3 but not L2. The lack of consistent, significant RC patterning in 264 individual mice in HC and HCO animals, suggests that spatial patterning within APC is not a reliable 265 feature of odor processing in familiar environments. In contrast, exploration of a novel odor environment 266 (NEO) strongly and reliably changes RC spatial patterning in L3 compared to HCO, and HC contexts. 267 This suggests that space may be an avenue to differentially process odor information depending on 268

context. 269
In the first section of this study, we show that inhibition of dPCs in L3 increases along the 270 rostral-caudal axis on the spatial scale of millimeters. We find that a disinhibitory circuit mediated by 271 SST-cells supports this gradient through rostrally-biased inhibition of interneurons ( Figure 6A). Could 272 these RC asymmetries in inhibition play a role in the RC patterning of neural activity in L3 during odor 273 exposure? In NEO animals, neural activity in L3 decreases from rostral to caudal APC over a spatial 274 scale of millimeters ( Figure 5D2) comparable to that of increasing inhibition in dPCs ( Figure 1F). In 275 Figure 6B, we plot the average normalized decrease in active L3 neurons in NEO mice and the 276 average normalized increase in inhibition of dPCs along the RC axis. We find that the spatial scales of 277 neural activity and inhibition are well matched with opposing slopes (NEO: slope -0.47 ± 0.04 mm -1 , R: 278 0.94, p<0.0001; Inhibition: slope 0.67 ± 0.06 mm -1 , R: 0.98, p: 0.0001). Thus, when inhibition is 279 weakest, neural activity is maximal (rostral) and when inhibition is strongest, neural activity is minimal 280 (caudal). These findings suggest that inhibitory circuitry could underlie RC patterning of neural activity 281 in L3 of APC. Further, the recruitment of inhibitory gradients and may depend on the context of odor 282 experience. In contrast, inhibition is weaker in L2 where the spatial profiles of neural activity are 283 approximately uniform in L2 and do not seem to vary with odor context. Altogether, these laminar and 284 RC differences in inhibition and neural activity suggest spatially dependent and independent 285 mechanisms work in parallel during odor processing in APC. 286

Discussion 287
In this study, we demonstrate rostral-caudal spatial patterning in inhibitory circuitry and neural 288 activity in APC. Our findings reproduce earlier studies that have shown caudally-biased asymmetric 289 inhibition of PCs 32 and a RC decline in fos(+) neurons following odor exposure 22 . However, the 290 underlying circuitry and functional significance of these findings are unknown. Here, we provide three 291 major advances. First, we describe a disinhibitory circuit mediated by SST-cells that decreases rostral 292 inhibition relative to caudal inhibition in L3 PCs. Second, we show that RC patterning of neural activity 293 is confined to L3 and differs with odor exposure in familiar (HCO) versus novel (NEO) contexts. Finally, 294 the density of active neurons decreases along the RC axis following odor exposure in the NEO context 295 commensurate with increasing inhibition of L3 PCs. Specifically, rostral PCs are more active and 296 receive significantly less inhibition (disinhibited) whereas caudal PCs receive stronger inhibition and are 297 less active. Altogether, our findings provide new evidence for RC spatial organization within APC as 298 well as a potential circuit mechanism for varying olfactory processing in different contexts. 299

Disinhibition by Somatostatin Interneurons 300
Inhibition plays a critical role in the processing and representation of sensory information in the cortex Recent studies have shown that SST interneuron activity is modulated in different contexts. In 313 sensory neocortices, SST-cell activity is enhanced by cholinergic modulation 56,60,61 , running during 314 visual sensory stimulation 62 , or engagement in an auditory task 61 . In somatosensory cortex, whisking 315 specifically increases the firing rates of fast-spiking, SST-cells that preferentially inhibit PV cells 56,59 . 316 Likewise, we have shown that two-thirds of SST cells in APC are FS and strongly inhibit PV-like cells 42 . 317 Given that sniffing and whisking are correlated 63,64 , an intriguing possibility is that actively exploring 318 (running, sniffing and whisking) a novel odor environment (NEO) globally enhances SST-cell activity in 319 sensory cortices. We propose that in APC, enhanced SST-cell activity gates rostral disinhibition and 320 increases in rostral neural activity in NEO animals. This interpretation is consistent with recent studies 321 that show interactions between interneurons in network models 65 promote context dependent changes 322 in network activity 58,61 . 323

Spatial patterning of neural activity in APC. 324
The spatial patterning is difficult to investigate in vivo due to the extent (~1.5 mm) and ventral location 325 of APC. Population imaging shows minimal spatial variation in response to different odors or intensities, 326 but typically only sample L2 neurons across ~300 µm of the RC axis 25-27 . Multi-site unit recordings 327 broadly sample the RC axis and suggest RC variation in odor-evoked firing rates 24 but sample a small 328 proportion of neurons per region. Likewise, intrinsic signal imaging or local field potential (LFP) 329 recordings broadly sample the RC axis and suggest systematic variation along the RC-axis in 330 concentration thresholds 28 and oscillatory activity respectively 13,29,30,66 . But these tools lack the fine 331 resolution to identify the neural circuits contributing to these responses. 332 To investigate the spatial profiles neural activity in vivo, we used TRAP-mice that conditionally 333 express of cre-recombinase linked to the IEG, c-fos 38 . This tool provides sufficient spatial scale to 334 investigate population activity along the entire RC axis at a resolution amenable microcircuit analysis. 335 TRAP-mice are advantageous over traditional IEG methods because cre-recombinase promotes 336 continuous cytoplasmic expression of tdTom independent of initial strength of activation. The limitation 337 is that the temporal window for capturing activity is longer. Labeling is optimal within one hour of 4-338 OHT injection and declines significantly ~6 hours post injection 38,67 . Thus, neural densities are 339 expected to be higher in TRAP animals due to enhanced labeling of weak responses and potential 340 spurious labeling over long time windows. To minimize the latter, animals were undisturbed in the dark 341 for 10h following exposure and HC animals provided a baseline for handling and non-specific labeling. 342 Consistent with previous IEG immunolabeling 22,68,69 , the density of activated, TRAP-tdTom(+) 343 cells increases significantly in odor-exposed animals (HCO and NEO) compared to HC animals. We 344 find these changes in density are restricted to L2 whereas RC patterning of neural activity occurs in L3. 345 A lack of RC patterning in L2 was initially surprising since L2 sPCs also receive caudally-biased 346 inhibition. However, L2 sPCs receive weaker inhibition than L3 dPCs 39 while L2 semilunar cells and 347 interneurons do not receive asymmetric inhibition. This suggests that the uniformly distributed spatial 348 pattern of neural activity in L2 is inherited from the spatial profile of afferent and/or recurrent excitation 349 18-20 . In contrast, interneuron densities, particularly SST cells, are greatest in L3 40,42 . We show that 350 individual L3 dPCs receive strong inhibition that increases with caudal position along the RC axis with 351 the same spatial scale as decreases in neural activity. These are ideal conditions for inhibition to 352 dictate L3 RC activity patterns. Altogether, laminar differences in inhibition and RC patterning coincide 353 with layer-specific differences PC subclasses and projection targets 70,71 and support the premise that 354 parallel processing streams exist in APC. 355

Functional roles for RC asymmetries in olfactory processing 356
Given the seemingly uniform profile of excitation in APC, a surprising finding is that inhibitory 357 strength increases along the rostral-caudal axis. In entorhinal cortex, a dorsal-ventral inhibitory gradient 358 coincides with an increase in PV interneuron density, changes in receptive field size and increased 359 gamma oscillatory power 72 . We find SST cells rather than PV cells change in density along the RC 360 axis. Since SST-cell inhibition has a subtractive effect on PC odor tuning 51 , it is possible that SST- were not distinguishable from noise and given a value of 0. To compare the spatial profiles of inhibition 428 across animals IPSC amplitudes were normalized to the strength of the maximum IPSC regardless of 429 location in the grid. The rostral-caudal bias was taken as the average normalized inhibition from the 430 caudal sites minus the average inhibition of the rostral sites, divided by the summed inhibition from both 431 sides. The bias metric ranges from -1 (rostral bias) to +1 (caudal bias). Since L1 inhibition was typically 432 weak 39 these sites were excluded from the bias metric.  Underlined values correspond to the cohort of mice exposed to ethyl butyrate (HCO, NEO only) and the 713 remaining mice were exposed isoamyl acetate. Linear regression was also performed on the average 714 normalized density across mice versus distance (Fit Average). P-values correspond to tests for slope 715 not equal to zero. Finally, the distribution of slope values was compared to zero using a non-parametric  is plotted against the difference in RC distance between the two cells. As the distance between the two 739 cells increases, the difference in inhibition also increases (* p<0.05, linear regression, n=19). F) 740 Inhibitory strength versus RC position of the dPC relative to the rostral start of the LOT in the sagittal 741 slice. Inhibition increases with RC distance (**p<0.01, linear regression, n=27). 742 average I C was significantly greater than I R (black ** p<0.01 paired t-test), while I R was significantly 755 greater than I C l in L3 INTs (green ** p<0.01, paired t-test). 756 The distribution of slope values was significantly non-zero (** p: 0.002 MWU-test). B1-4) As in A1-4, 764 except for Calbindin(+) interneurons (CB). B1,2) Data from CB mouse #5 in Table 1.

B3) On average, 765
there is no change in density of CB cells along the RC axis (filled green circles, p: 0.55). B4) In 766 individual mice CB cells significantly increased or decreased (filled green circles) along the RC axis, but 767 the distribution of slopes did not differ from zero (p: 0.37, MWU-test). C1-4) As in A1-4, except for 768 Parvalbumin(+) interneurons (PV). B1,2) Data from PV mouse #6 in Table 1. C3) On average, the 769 density of PV cells decreased along the RC axis (filled black circles, p: 0.03). C4) However, only one 770 mouse showed a significant decrease in PV cells along the RC axis (filled black circle) and the 771 distribution of slopes did not differ from zero (p: 0.07, MWU-test). 772 to density versus distance for individual mice in each condition in L2 (E1) and L3 (E2). Filled triangles 799 indicate slopes significantly different from zero (p<0.05, see Table 1 for p: values). E2) In L3, the 800 distribution of slopes in HCO and NEO animals significantly differed from 0 (** p<0.01, MWU-test). 801 Further the distribution of slopes in NEO animals significantly differed from HC and HCO animals (★★ 802 p<0.01, KW-test). F) The average density of neurons increases with odor exposure (HCO, NEO) in L2 803 (F1, * p<0.05, KW-test) but not L3 (F2). 804  Table 1, ** p<0.01. A3) Normalized density of SST-cells versus distance for all mice (open circles). The average (± SE, n=7) normalized density of SST cells across mice significantly decreased with RC distance (red circles, **p<0.01). A4) Distribution of slopes from linear fits to data from individual mice. Solid circles indicate significantly negative slopes (p<0.05). The distribution of slope values was significantly non-zero (** p: 0.002 MWU-test). B1-4) As in A1-4, except for Calbindin(+) interneurons (CB). B1,2) Data from CB mouse #5 in Table 1. B3) On average, there is no change in density of CB cells along the RC axis (filled green circles, p: 0.55). B4) In individual mice CB cells significantly increased or decreased (filled green circles) along the RC axis, but the distribution of slopes did not differ from zero (p: 0.37, MWU). C1-4) As in A1-4, except for Parvalbumin(+) interneurons (PV). B1,2) Data from PV mouse #6 in Table 1. C3) On average, the density of PV cells decreased along the RC axis (filled black circles, p: 0.03). C4) However, only one mouse showed a significant decrease in PV cells along the RC axis (filled black circle) and the distribution of slopes did not differ from zero (p: 0.07, MWU). Mice continuously move between the odorized (orange circle (+)) and non-odor arms (blank circle (-)) as well as the center (C) of the arena. C, D) Normalized density of tdTom(+) cells along the RC axis of L2 (C) and L3 (D). Open triangles: data from individual animals; filled triangles: average across animals. E) Slopes of linear fits to density versus distance for individual mice in each condition in L2 (E1) and L3 (E2). Filled triangles indicate slopes significantly different from zero (p<0.05, see Table 1 for p: values). E2) In L3, the distribution of slopes in HCO and NEO animals significantly differed from 0 (** p<0.01, MWU-test). Further the distribution of slopes in NEO animals significantly differed from HC and HCO animals ( p<0.01, KW-test). F) The average density of neurons increases with odor exposure (HCO, NEO) in L2 (F1, * p<0.05, KW-test) but not L3 (F2).

Rostral Caudal
Rostral Caudal A2 Figure 6: The spatial profiles of inhibition are commensurate with neural activity along the RC axis of APC. A1) Schematic summarizing rostral-caudal spatial profiles of inhibition (green), SST-interneuron density and SST-mediated inhibition of interneurons (blue) and active neurons in L3 of NEO mice (red). A2) Proposed disinhibitory circuit consisting of a higher density of rostral SST-cells (blue) that inhibit interneurons (light green) and disinhibit PCs increasing rostral neural activity (red PCs). In caudal APC, lower density of SST-cells allows greater inhibition (dark green) and less active PCs (black). B) Normalized density of active neurons (red triangles) decreases along the RC axis (slope: -0.47 mm-1, R 2 : 0.94, p<0.0001, from Figure 5D3), as normalized inhibition of dPCs increases (slope: +0.67 mm-1, R 2 : 0.95, p<0.0001, from Figure 1F). Because there are only two points at 1.5 mm in Figure 1F, inhibitory strength was normalized to a projected "maximum" strength filled triangles) versus distance along the RC axis of Fos-tdTom(+) cells for the mouse corresponding to the sections above. Linear regression fits (thin lines, L2: thick lines, L3) with slopes that significantly differed from zero are indicated by astrisks: * p<0.05, or ** p<0.01. Each column is data from an individual mouse and the mouse number (M1-6) matches the mouse numbers in      Supplemental Table 2: Summary stats for F-test for non-zero slope of linear regression. nnumber of samples, "1-β" power analysis results at α: 0.05 given sample number. Bold: Significant p-values, red : 1-β < 0.7. For density measures, power was analyzed for the mice yielding significant results but the lowest R-values and the lowest number of samples. These represent the minimum power for significant findings within the group.

Figure 5
Density