Organization and engagement of a prefrontal-olfactory network during olfactory selective attention

Sensory perception is profoundly shaped by attention. Attending to an odor strongly regulates if and how a smell is perceived – yet the brain systems involved in this process are unknown. Here we report integration of the medial prefrontal cortex (mPFC), a collection of brain regions integral to attention, with the olfactory system in the context of selective attention to odors. First, we used tracing methods to establish the tubular striatum (TuS, also known as the olfactory tubercle) as the primary olfactory region to receive direct mPFC input in rats. Next, we recorded local field potentials from the olfactory bulb (OB), mPFC, and TuS while rats completed an olfactory selective attention task. Gamma power and coupling of gamma oscillations with theta phase were consistently high as rats flexibly switched their attention to odors. Beta and theta synchrony between mPFC and olfactory regions were elevated as rats switched their attention to odors. Finally, we found that sniffing was consistent despite shifting attentional demands, suggesting that the mPFC-OB theta coherence is independent of changes in active sampling. Together, these findings begin to define an olfactory attention network wherein mPFC activity, as well as that within olfactory regions, are coordinated in manners based upon attentional states.


Introduction 18
Sensory processing and thus perception are both profoundly shaped by our ever-19 changing cognitive states. In most cases, the thalamus appears to be a major driver of 20 state-dependent modulation of sensory information (Wimmer et  Kay and Sherman 2006). Thus, other brain systems must engage with olfactory 31 processing in order to afford one the ability to attend to odors. This is a significant issue 32 since odors are most often encountered in highly multisensory environments, for instance 33 during eating, wherein potentially distracting or conflicting cues must be ignored at the 34 expense of selectively attending to odor. 35 Truly very little is known regarding the neural mechanisms underlying olfactory 36 attention. One brain region that seems likely to confer this ability, at least in part, is the 37 tubular striatum (TuS, also known as the olfactory tubercle (Wesson 2020)). This is true 38 in both humans and rodents. For instance, early work using fMRI uncovered the first 39 evidence that the human TuS is more activated in response to attended versus

159
Among PFC subregions, layer 5 PrL and IL neurons provide the densest input to the TuS. 160 While the PrL and IL are strongly implicated in attention, the PFC also includes the OFC. 161 The OFC is involved in polysensory processing (Rolls 2004 quantified except the IL and the MO (Fig. 2D). Thus, the PrL and IL provide the densest 173 inputs to the TuS and are well-positioned to influence odor processing. 174 Within the PrL and IL, we quantified cell body locations throughout the cortical 175 layers for both the ipsilateral and contralateral hemispheres. We observed that in both of 176 these regions, the majority of cell bodies were found in layer 5 (Fig. 2E) (Fig. 3). 209 Briefly, the CAT is a modified two-alternative choice task in which rats are simultaneously 210 presented with one of two olfactory cues (odor A/odor B) and one of two auditory cues 211 (tone on/tone off). A single behavioral session begins with tone attention: that is, the tone 212 cues signal the reward port location whereas the odor cues are distractors ( Fig. 3A-C, 213 blue shading). Once criterion on the tone attention phase of the task has been reached 214 (6 blocks of 20 trials at ≥80% correct), an uncued intermodal rule change occurs, and the 215 rats must now direct their attention to odors, and ignore tones, to accurately locate their 216 rewards ( Fig. 3A-C, orange shading). This rule change is accompanied by a temporary 217 drop in performance as the rats adjust their behavior to the new rule (Fig. 3C, pink  218 shading) before they eventually perform well on odor attention (6 blocks at ≥80% correct, 219 orange shading). We will refer to the blocks following the rule change before performance 220 reaches ≥80% correct on odor attention as "switch" blocks, in which the rat is by-definition 221 performing poorly. Importantly, each session began and ended with 3 blocks of odor-only 222 trials, in which there were no competing tone cues, to use as a control for odor 223 discrimination without any additional cognitive demand. 224 In the CAT, there are four possible combinations of trials (Fig. 3B). Two of these 225 are "congruent," in that the olfactory and auditory cues signal approach to the same 226 reward port, and two are incongruent, in that the cues signal opposite ports. Thus, the 227 correct reward port on incongruent trials depends on the current task rule: tone attention 228 (blue shading) or odor attention (orange shading). Importantly, all analyses of 229 physiological signals were limited to trials on which the tone was off to avoid multisensory 230 influences and focus on the effects of cognitive state on odor processing specifically 231 (Carlson et al. 2018). On a single trial of the CAT, the rat will nose poke to initiate a trial, 232 then must hold in the center port for 1 second before the stimuli come on (Fig. 3D, dark 233 gray shading). Then, the odor and tone stimuli come on simultaneously, and the rat must 234 remain in the center port for at least 400 ms sampling the stimuli (Fig. 3D, light gray 235 shading). After 400 ms, the rat is free to make a choice at the left or right port and receive 236 a water reward if correct. 237 We simultaneously recorded LFPs from the TuS, mPFC, and OB from 5 highly-238 proficient expert rats (see Methods) while they performed the CAT (Fig. 3D-E). This 239 allowed us to explore network dynamics locally within each structure, as well as coherent 240 activity between these structures.

245
These cues direct the rat to retrieve a fluid reward at either the left or the right port (outcome).

246
Behavioral sessions begin with tone attention (auditory cues predict reward; blue shading) and    Hz) and high (60-80 Hz) gamma range within each brain structure as rats completed the 270 CAT (Fig 4). To do this, we measured the power of gamma oscillations within the hold 271 and odor trial epochs across each task type (odor only, tone attention, switch, and odor 272 attention), and normalized these values to those for odor only trials to highlight the specific 273 contributions of sensory-directed attention as compared to 'basic' olfactory discrimination. 274 In the TuS, we observed a slight enhancement in low gamma oscillations with 275 increased attentional demand, but this did not reach statistical significance across rats 276 (Fig. 4B). Interestingly, we observed that high gamma oscillations in the mPFC were 277 elevated during odor attention compared to tone attention (Fig. 4B). We were surprised 278 to observe this enhancement specifically for odor-directed attention in the mPFC, since 279 one might anticipate increased gamma power with increased cognitive demand 280 regardless of sensory modality. In the OB, we observed increased power of low gamma 281 oscillations during odor attention as compared to odor only, indicating that increased 282 attentional demand alone is enough to modify odor information at the earliest stage of 283 processing in the brain (Fig. 4B). Finally, OB oscillations in the high gamma range were 284 elevated in power during the attentional switch relative to odor only and tone attention, 285 suggesting network activity related to cognitive flexibility. While we did uncover some 286 changes in beta band power (Fig. S2), these were not as dramatic across attentional 287 states as was the case with gamma. Overall, these findings indicate changes in local 288 network dynamics in the OB and the mPFC during selective attention to odors, suggesting 289 that attention may modulate odor processing at its most early processing stage (the OB).

295
Quantification of power in the low and high gamma ranges across all task types, normalized to 296 odor only. For each region/frequency band, a 2-way ANOVA with Geisser-Greenhouse correction 297 was completed. TuS, Low gamma: main effect of task type, F(1.62, 6.5)=6. 04

304 305
Olfactory bulb gamma oscillations couple with theta phase during selective attention. 306 In some brain regions, the amplitude of high frequency oscillations are structured by the 307 phase of low frequency oscillations, a phenomenon known as phase amplitude coupling 308 To address this, we first computed comodulograms to identify high frequency 316 oscillations coupled to theta phase within the rat OB during the CAT, which revealed 317 strong coupling between theta and high gamma (Fig 5A), and much weaker coupling 318 between theta and beta (Fig S3). To investigate the significance of theta-high gamma 319 PAC, we examined the trial-by-trial amplitude of high gamma power as a function of theta 320 phase, which indicated high coupling throughout individual sessions and across cognitive 321 states ( Fig. 5B-C). Peak phase angle was consistent even comparing correct vs. incorrect 322 trials (Fig. 5D). For each task type, we computed the modulation index (MI) of the PAC, 323 which is a measure of the extent to which a given high-frequency oscillation is structured

355
Beta oscillations are more coherent between the mPFC and olfactory regions during an 356 intermodal attentional shift. 357 We next tested whether spectral activity between the mPFC and olfactory system might 358 become more coherent during attention. We observed enhanced coherence in the beta 359 range (15-35 Hz) between the mPFC and the TuS specifically during the switch blocks -360 when the rule has been changed from tone to odor attention, but the rats have not yet 361 successfully switched their attention (Fig. 6A). For the mPFC-TuS, this elevation was 362 specific to the 1 second hold period prior to odor onset (Fig. 6B) which corresponds to 363 anticipation. Between the mPFC and the OB, we similarly observed increased coherence 364 in the beta band, but during both the hold and odor epochs (Fig. 6C-D). In contrast, no 365 changes in coherence between the OB and TuS were uncovered (data not shown). 366 Overall, these data indicate that mPFC engagement with olfactory structures is 367 upregulated during a cognitively demanding switch from auditory to olfactory selective 368 attention, suggesting a role for the mPFC in attention-dependent odor processing. Fontanini and Bower 2006). This is particularly relevant in an odor-guided task, where 390 correct performance depends upon sampling of the odors via sniffing. We observed that 391 during the attentional switch, there was a striking increase in coherence in the theta range 392 (2-12 Hz) compared to the odor attention state (Fig. 7A-B). While a slight increase was 393 observed during the hold epoch (Fig. 7A-B), this increase was much more pronounced 394 and statistically significant during the odor sampling period (Fig. 7A-B). While our prior 395 analyses had been restricted solely to correct trials for odor only, tone attention, and odor 396 attention, we included correct and incorrect trials for all switch blocks, since this switch 397 state is defined by poor performance and behavioral flexibility, and also because this 398 allowed for the inclusion of comparable numbers of trials in the analysis (see Methods). 399 Thus, we separated trials for switch blocks only into correct and incorrect trials, discarding 400 a random selection of correct trials to match the number of incorrect trials available. This 401 revealed, counterintuitively, a greater coherence on incorrect compared to correct trials 402 (Fig 7C), suggesting that OB-mPFC theta band coherence is upregulated in contexts 403 where behavioral flexibility is required.   this, we trained a separate cohort of rats to perform the CAT before implanting 435 thermocouples in their nasal cavities, allowing us to monitor sniffing behavior by 436 measuring temperature changes (airflow) within the nasal cavity (Fig. 8A-B). Unlike  Kepecs et al. 2007). Therefore, we quantified sniffing 441 frequency specifically during the hold and odor periods. As illustrated by the example 442 session from one rat in Fig. 8C, we observed no clear changes in sniffing behavior across 443 task types (Fig. 8C-D). Instead, the rats displayed a highly stereotyped pattern of sniffing 444 behavior, suggesting that reaching high proficiency on the CAT results in their 445 development of a sensorimotor program that is implemented on each trial, regardless of 446 current attentional demand (Fig. 8C, bottom). This was the case across all rats. Although 447 we observed a slight decrease in sniffing frequency specifically during the hold period 448 25 throughout a session on average (Fig. 8D, F), this was confined to the hold period as the 449 rats anticipated odor arrival and sniffing frequency during the odor sampling period 450 remained remarkably constant throughout the sessions for 2/3 rats (Fig. 8D, F). Given 451 the lack of changes in sniffing frequency by task type, we investigated whether the timing 452 of sniffs during the odor period were more intentional in relation to odor onset when 453 animals were faced with attending to odor. We examined the time to the first sniff across 454 task types and found no difference, suggesting that sniff timing relative to odor onset is 455 independent of attentional demands (Fig. 8E). We also examined sniffing frequency on 456 correct versus incorrect trials during the switch blocks (Fig. 8G) yet did not identify 457 differences in sniffing frequency, providing further evidence that changes in sniffing 458 behavior do not likely account for modulations in OB-mPFC coherence that we 459 uncovered, which were correlated with trial outcome. Together, these data indicate that 460 sniffing strategies in rats are resilient to enhanced attentional demand, providing evidence 461 for covert (rather than overt) olfactory attention in rodents.

485 486
Discussion 487 Here we used anatomical, behavioral, and physiological approaches to demonstrate 488 integration of the mPFC with the olfactory system in the context of selective attention to 489 odors. We show that mPFC neurons in the PrL and IL subregions directly and 490 preferentially target the mTuS compared to other olfactory regions, suggesting that they 491 are well-positioned to exert influence on olfactory processing via the mTuS. We then used 492 a physiological and behavioral approach to demonstrate local and interregional effects of 493 attention on network activity within and between the mPFC and olfactory regions, 494 including the OB and TuS. Finally, we found that olfactory sampling behavior is resilient 495 to attentional demand, indicating that olfactory attention may be an "covert" rather than 496 "overt" process. Together, this work adds to a growing body of literature on the possible 497 mechanisms underlying cognitive modulation of olfactory processing and thus perception. 498 499 Insights into mPFC connectivity with the olfactory system. 500 Our tracing experiments uncovered previously unappreciated aspects of mPFC 501 connectivity with the olfactory system. We found that the PrL and IL most densely 502 innervate the mTuS compared with other olfactory regions. By using a combinatorial AAV 503 approach, where Cre expression driven by the CaMKII promotor permits expression of 504 synaptophysin-eGFP/-mRuby, we were able to identify this pathway as excitatory while 505 confidently attributing fluorescence in the TuS (and PCX) to synaptic terminals (primarily 506 in layers 2 and 3) rather than fibers of passage (Fig. 1). We further demonstrated that 507 among PFC subregions, the PrL and IL provide the most projections to the TuS, with the 508 MO coming in third, and these projection neurons mostly reside in layer 5 (Fig. 2). Our 509 data are in agreement with earlier tracing work which established that rat mPFC neurons 510 project throughout the brain, including in the TuS (Vertes 2004), and a recent review 511 proposing that the PrL, IL and MO be grouped together as the ventromedial PFC, based Our multisite LFP recordings during attentional performance uncovered many changes in 526 network activity which expand our appreciation for how the olfactory system is shaped by 527 cognitive state. There are several especially notable outcomes we discuss here. 528 First, while we know that the mPFC is crucial for attention, no studies have 529 monitored mPFC network activity during olfactory attention, leaving a major void in our 530 29 understanding of how the mPFC engages with the olfactory system. Because the mPFC 531 is integral for some forms of attention, we predicted it may be recruited during olfactory 532 attention. In support of this, we observed elevated gamma power in the mPFC during 533 odor-directed attention relative to tone attention (Fig. 4). The mPFC is certainly not an 534 olfaction-specific structure, though it is engaged by odor-guided tasks requiring learning 535 (Wang et al. 2020) and high working memory capacity (De Falco et al. 2019). This 536 elevation in gamma power does not likely reflect increased reward confidence, as 537 behavioral performance was comparable across task types (Fig. 3C). It is interesting to 538 consider whether the mPFC of rodents is predisposed to favor and prioritize olfactory 539 information more so than other sensory stimuli. Nevertheless, these findings exhibit 540 engagement of the mPFC during odor-directed attention, providing support for its 541 inclusion in an olfactory attention network. 542 543

Olfactory attention enhances power of OB gamma oscillations. 544
Our work is the first to monitor OB activity during selective attention. We found that 545 attention powerfully shapes OB activity, which implies that odor information received by 546 structures downstream from the OB, including the TuS, is subject to attention-dependent 547 modulation. Specifically, we observed increased power of low gamma oscillations (40-60 548 Hz) in the OB during odor-directed attention as compared to odor only discriminations 549 (Fig. 4). Additionally, we observed elevated power of high gamma oscillations (60-80 Hz) 550 while rats attempted to switch their attention from tones to odors (Fig. 4). While increased 551 gamma power in the OB has been associated with successful discrimination of 552 perceptually similar vs. dissimilar odors (Beshel et al. 2007), our findings indicate that 553 similar effects can be observed when the odor discrimination is simple/coarse, but the 554 attentional demand is high. Interestingly, elevated low gamma power during odor 555 attention was evident during both the hold and odor epochs, while elevated high gamma 556 power during switch blocks was isolated to the odor sampling period (Fig. 4B). High and 557 low gamma oscillations are considered distinct phenomena in the OB, and are believed 558 to have mechanistically unique origins (Kay 2003), so it is perhaps not surprising to 559 observe modulation of these frequency bands during different attentional demands. Low 560 gamma oscillations are believed to arise from inhibition between local interneurons, are 561 unstructured relative to the sniff cycle, and are functionally mysterious, though they have 562 been observed in states of engaged quiescence (Kay 2003). Our data thus support a 563 potential role for low gamma oscillations in attentionally demanding odor discriminations, 564 though future work is needed to fully appreciate the mechanisms of this. 565 In contrast, high gamma is structured to the sniff cycle, and is generated by local is consistent with these mechanistic underpinnings and suggests specific changes in the 576 nature of odor processing as one undergoes a cognitively demanding switch to odor 577

attention. 578
As mentioned, high frequency gamma in the OB is consistently aligned with the 579 respiratory cycle, which was evident in our PAC analysis (Fig. 5). Recent work 580 demonstrated that OB theta-high gamma PAC is strengthened as mice learn to 581 discriminate odors in a go-no go task, specifically for the go stimulus, suggesting that 582 PAC may support olfactory behavior (Losacco et al. 2020a) and leading us to test whether 583 attention employs (or perhaps just simply influences) OB PAC. Our results uncovered 584 highly consistent theta-gamma PAC in the OB across attentional demands, and much 585 weaker coupling between theta and beta oscillations, leading us to focus on theta-gamma 586 PAC. However, we did not observe a decrease in PAC when expert rats completed trials 587 incorrectly (Fig. 5D), suggesting that perhaps PAC is not necessary to successfully 588 discriminate coarse odor pairs, like the ones we used herein. One possible explanation 589 for this difference is that in 2-alternative choice tasks, like the CAT, both stimuli are 590 assigned positive valence, while in go no-go tasks like that used by (Losacco et al. 591 2020a), one stimulus loses positive valence upon learning. Throughout a single session 592 of the CAT, odors temporarily lose their reward-predictive value during tone attention, but 593 it is quickly regained (e. g., Fig. 3C). Our data indicate that OB PAC, in rats who have 594 been shaped to expert level on the same odor discrimination over many weeks, is resilient 595 to a temporary lapse in positive odor valence, and perhaps supports flexible behavior 596 supporting attentional switches. 597 598 599 32 Beta synchrony integrates mPFC activity within the olfactory network. 600 Our data are the first to show functional coupling between the mPFC and olfactory regions 601 during attentional demandsspecifically during an intermodal attentional shift to odors 602 (Fig. 6) TuS as rats attempted to switch their attention from tones to odors (Fig. 6A-B). In the 607 context of our finding that the mPFC and the TuS are connected via a unidirectional 608 monosynaptic pathway (Figs. 1-2), these data suggest that communication between the 609 mPFC and TuS is strengthened during attentional shifts. Indeed, ventral striatum- to demonstrate that this concept is applicable to the olfactory system, which possesses 617 unique anatomical organization. 618 In addition to enhanced mPFC-TuS coherence, we also observed enhanced beta 619 band coherence between the mPFC and the OB (Fig. 6C-D), raising the intriguing 620 possibility that prefrontal influence on olfactory processing could begin as early as the 621 OB. While the OB and mPFC are not connected monosynaptically (Fig 1) OB and mPFC (Fig. 7), that enhanced attentional demand may influence sampling 658 strategy. For instance, a rat might increase sniffing frequency during odor sampling when 659 attention to odors versus attending to tones. We found that rats' sniffing strategies were 660 remarkably resilient to shifting attentional demands, remaining stereotyped as rats flexibly 661 switched their attention from the auditory to olfactory modality (Fig. 8). This is in contrast 662 to some findings in humans, indicating that humans alter the timing and depth of their It is interesting to consider sensory sampling via sniffing as analogous to saccadic 671 eye movements (Uchida et al. 2006), which contribute to rhythmic attentional sampling in 672 the visual system (Fiebelkorn and Kastner 2019; VanRullen 2016). In the visual system, 673 attention is regarded as overt when it is accompanied by saccadic eye movements to a 674 target and covert when the eyes remain fixated on a central point (Posner et al. 1980). 675 The investigation of these different modes of attention and their underlying networks has 676 spanned decades (Posner 2016). While these two processes engage similar brain 677 networks (Rizzolatti et al. 1987;Corbetta 1998), suggesting that they may not actually be 678 separate, other work suggests different populations of neurons within these networks may 679 support each type of attention (Thompson et al. 2005). Analogously, our observation of interregional coupling between the mPFC-OB and mPFC-TuS. Our data suggest that 691 changes in sniffing do not drive these effects, highlighting that odor-directed attention, at 692 least in this context, is orchestrated by top-down mechanisms, as opposed to 'bottom-up' 693 influences (from odor sampling). Overall, these findings begin to reveal an olfactory 694 attention network and bring us closer to understanding how the brain affords the ability to 695 selectively attend to odors. 696

Materials and Methods 697
Animals 698 Adult, male Long-Evans rats were obtained from Charles River (Wilmington, MA) and 699 Envigo (Indianapolis, IN) and maintained in the University of Florida vivarium on a 12:12 700 light:dark cycle, with food and water provided ad libitum until water restriction for 701 behavioral shaping began. All experiments were conducted in accordance with NIH 702 guidelines and were approved by the University of Florida Institutional Animal Care and 703 Use Committee. 704 705

Surgical procedures 706
For all surgical procedures, rats were maintained on 4-1% isofluorane in 1.5 L/min O2 and 707 placed in a stereotaxic frame. The scalp was shaved and cleaned with betadine and 70% 708 ethanol. Analgesia in the form of meloxicam was administered (5 mg/kg s.c.) and the local 709 anesthetic marcaine (5 mg/kg s.c.) was given prior to the cranial incision. A cranial incision 710 was made and the skin was retracted using hemostats. 711 For viral injections, a craniotomy was then drilled over the region of interest, and a 712 glass micropipette containing AAV was slowly lowered into region of interest. For 713 anterograde mPFC injections (Fig. 1) Watertown, MA; 105558-AAV9, titer 1x10^13 vg/mL) was injected into the IL, then the 718 PrL, at a rate of 2 nL/sec. For retrograde mPFC injections (Fig. S1), 200 nL total of AAVrg-719 hSyn-GFP (Addgene 50465-AAVrg; titer 7x10^12 vg/mL) was unilaterally injected at a 720 rate or 2 nL/sec into the mPFC (100 nL in IL, followed by 100 nL in PrL). For TuS 721 injections, 200 nL of AAVrg-hSyn-GFP (Addgene 50465-AAVrg; titer 7x10^12 vg/mL) was 722 injected unilaterally at a rate of 2nL/sec. In all cases, after waiting 5 minutes, the pipette 723 was slowly withdrawn from the brain, the craniotomy was sealed with dental wax, and the 724 incision was sutured. 725 For electrode implants, the skull was scrubbed with 3% H2O2 and covered with a 726 thin layer of cyanoacrylate (Vetbond, 3M). Craniotomies were drilled over each brain area 727 of interest, plus 3 craniotomies for 0-80 stainless-steel screws to aid in anchoring the 728 dental cement. After drilling craniotomies over each brain area of interest, the bipolar 729 stainless-steel electrodes (0.005-in outer diameter, Teflon coated to 0.007-in outer 730 diameter) were lowered into the brain and secured with a small amount of dental cement 731 before moving on to the next electrode. Once all wires were placed and secured, an 732 electrical interface board (EIB) (Open Ephys, Cambridge, MA) fitted with a 32 channel 733 connector (Omnetics, Minneapolis, MN) was lowered over the skull, and the electrode 734 wires were secured to the desired channels using gold pins. After the stainless-steel 735 ground wire was secured to a skull screw with conductive silver paint, the whole assembly 736 was secured with dental cement. 737 For thermocouple implants (Wesson 2013; Uchida and Mainen 2003), following 738 skull preparation as above, a craniotomy was made in the nasal bone (0.9 mm lateral 739 from midline) and a thermocouple wire was lowered 3 mm into the nasal cavity and 740 secured with dental cement. Then, as for the electrode implants, an EIB with a 32-channel 741 Omnetics connector was lowered over the skull, the thermocouple leads secured with 742 gold pins, and a ground wire secured to a skull screw before the whole assembly was 743 secured with dental cement. 744 Following surgery, rats were returned to their home cages to recover on a heating 745 blanket. The rats received post-operative analgesia for at least 3 days mixed with a 746 palatable gel (5 mg/kg meloxicam in Medigel, ClearH2O, Westbrook, ME). Electrode 747 implanted rats were implanted prior to the beginning of behavioral shaping. Thermocouple 748 implanted rats were shaped prior to surgery and were allowed full water access for at 749 least 24 hours prior to surgery. All rats were allowed to recover for at least 5 days before 750 beginning or restarting water restriction. 751 752

Perfusion and histology 753
For anterograde mPFC viral injections (Fig. 1), rats were perfused 2-4 weeks following 754 injection. For retrograde mPFC (Fig. S1), and TuS viral injections (Fig. 2), rats were 755 perfused 2 weeks following injection. All rats were overdosed with Fatal-Plus and 756 perfused with cold 0.9% NaCl followed by cold 4% formalin. Brains were dissected and 757 stored in 10% formalin in 30% sucrose prior to sectioning. Alternate 40 um sections were 758 collected with a sliding microtome and stored in Tris-buffered saline with 0.03% sodium 759 azide. For electrode implanted rats, sections were mounted on gelatin subbed slides and 760 stained with 0.1% cresyl violet to confirm electrode locations. 761 762

Image acquisition and quantification 763
Brain areas of interest were identified using the rat brain atlas (Paxinos and Watson 764 1997). Images were acquired with a Nikon Eclipse Ti2e fluorescent microscope at 20x 765 magnification using a Nikon 16 MP DS-Qi2 monochrome CMOS camera. For all tracing 766 experiments, successful targeting of the desired subregion was confirmed, and injections 767 with spillover into surrounding regions were excluded. For anterograde mPFC injections 768 (Fig. 1), if one of the two injections was on target, we analyzed only that region and 769 disregarded the other. Overall, we analyzed 8 rats with on target PrL injections and 5 rats 770 with on target IL injections, with 3 rats having both PrL and IL quantified. From these rats, 771 images for quantification were acquired as follows: for the TuS, 11 images per rat, evenly 772 spanning 2.7mm anterior -0.8 mm posterior Bregma. For the PCX, 17 images per rat 773 were quantified, evenly spanning 3.7mm anterior -4.8mm posterior Bregma. For the 774 AON, 6 images per rat were quantified, evenly spanning 5.7mm-2.7mm anterior Bregma. 775 For retrograde TuS injections (Fig. 2), 3-10 (6.46±2.48) PrL/IL-containing sections and 1-776 6 (3.9±1.46) OFC-containing sections were imaged (n = 6 rats). 777 After acquiring images, ROIs were drawn around each area of interest and 778 fluorescent puncta or cell bodies were detected using semi-automated counting 779 algorithms created within NIS elements software (Nikon) based on their fluorescence 780 intensity and size. Cell or puncta counts were then normalized to the ROI area for 781 comparison across regions. For layer-specific quantification (Fig. 2E), custom MATLAB 782 code was used to determine the layer in which each counted cell resided. The medial and 783 lateral TuS were defined as the medial and lateral third of the TuS, to ensure clear 784 separation between the regions. In puncta quantification, we initially differentiated 785 between the anterior and posterior PCX, which was divided based on the presence or 786 absence respectively, of the lateral olfactory tract. Because we observed no differences 787 in puncta between the anterior and posterior PCX for PrL (paired, two-tailed t-test, p=0.32) 788 examine coupling with theta frequencies down to 2 Hz. Still, this time segment allows for 880 the analysis to be contained to an individual trial without any overlap with neighboring 881 trials. 882 883

Sniffing analysis 884
As with the LFP analyses, all analyses on the sniffing data were restricted to trials where 885 the tone was off. Additionally, only correct trials were examined for odor only, tone 886 attention, and odor attention blocks, while correct and incorrect trials were examined for 887 switch blocks. The data were imported to MATLAB and traces spanning -10 to 8.4 sec 888 from odor onset from each trial were filtered using a 2 nd order band-pass filter from 0.5-889 10 Hz. After extracting trials as described above for the LFP analyses, these filtered traces