CA1 20-40 Hz oscillatory dynamics reflect trial-specific information processing supporting nonspatial sequence memory

The hippocampus is known to play a critical role in processing information about temporal context. However, it remains unclear how hippocampal oscillations are involved, and how their functional organization is influenced by connectivity gradients. We examined local field potential activity in CA1 as rats performed a complex odor sequence memory task. We find that odor sequence processing epochs were characterized by increased power in the 4-8 Hz and 20-40 Hz range, with 20-40 Hz oscillations showing a power gradient increasing toward proximal CA1. Running epochs were characterized by increased power in the 8-12 Hz range and across higher frequency ranges (>24 Hz), with power gradients increasing toward proximal and distal CA1, respectively. Importantly, 20-40 Hz power increased with knowledge of the sequence and carried trial-type-specific information. These results suggest that 20-40 Hz oscillations are associated with trial-specific processing of nonspatial information critical for order memory judgments.

, and that distinct oscillatory states are observed in hippocampal subregion CA1 52 during spatial navigation in rodents (see Colgin, 2016). In addition to the prominent theta rhythm 53 (8-12 Hz), there is also evidence that the CA1 network exhibits transient increases in slow 54 gamma (25-55 Hz) and fast gamma (60-100 Hz) power during running. In fact, a landmark study 55 by Colgin and colleagues (2009) showed that CA1's slow gamma oscillations are coherent with 56 CA3 activity, whereas CA1's fast gamma oscillations are coherent with entorhinal activity, 57 suggesting these brief oscillatory states reflect retrieval and encoding processes, respectively. 58 However, establishing a direct link between these oscillatory states and specific forms of 59 information processing remains challenging, as such paradigms tend to have poor control over 60 the timing at which information is encoded or retrieved. the proximodistal axis of CA1 (higher power in distal than proximal CA1; Igarashi et al., 2014). 77 However, it remains unclear whether these oscillations are associated with specific cognitive 78 processes or aspects of performance, and whether proximodistal gradients in oscillatory power reflect the modality of the stimulus or vary with task demands. Further, it remains to be 80 determined whether oscillations observed during spatial exploration extend to the processing of 81 nonspatial information. 82 83 To address these important issues, we examined local field potential (LFP) activity in CA1 as 84 rats performed a hippocampus-dependent odor sequence memory task (Fig 1). Importantly, this 85 complex task offers precise time-locking to stimulus presentations and responses, as well as 86 distinct trial types, contrasts and time windows associated with distinct cognitive demands. As in 87 our previous work (Allen et al., 2016), we report prominent oscillations in the 20-40 Hz and 4-8 88 Hz frequency ranges during the odor sequence processing periods. Here we extend these 89 results by demonstrating that the same electrodes exhibited a distinct spectral content in a 90 different state (running on the track), which was characterized by high power in the 8-12 Hz 91 band and a broad but modest increase in power for frequencies above 24 Hz. In both behavioral 92 states, the power of recruited oscillations was found to vary along the proximodistal axis of CA1. 93 We also made two additional contributions to our understanding of 20-40 Hz oscillations to 94 hippocampal function. First, we discovered that 20-40 Hz oscillations were linked with sequence 95 memory performance, whereas oscillations in other frequency ranges did not show a significant 96 association. 20-40 Hz power increased with knowledge of the odor sequence, suggesting this 97 signal is associated with learning, and was differentially recruited across trial types, offering 98 strong evidence for its behavioral relevance. Second, we found that 20-40 Hz power was higher 99 in proximal than distal CA1, which is the opposite pattern to that observed in an odor-place 100 association task (Igarashi et al., 2014), suggesting that oscillatory power gradients along the 101 proximodistal axis may reflect task-specific demands. In light of prior evidence that proximal 102 CA1 is strongly associated with the medial entorhinal cortex (MEC; van Strien, Cappaert, and 103 Witter, 2009; Witter et al., 2017), and that MEC inactivations impair temporal coding in CA1 104 (Robinson et al., 2017), this finding suggests that functional coupling between proximal CA1 and 105 MEC may play a key role in remembering the temporal context of nonspatial events.

107
Oscillatory states differ between odor sequence processing and running periods and 108 vary across the proximodistal axis of CA1 109 We began by investigating whether there are distinct oscillatory states in CA1 that are unique to 110 the odor sequence processing component of the task, and whether the observed oscillations 111 varied across the proximodistal axis of CA1. To do so, group (n=5) peri-event spectrograms 112 were generated from four electrode locations along the proximodistal axis and aligned to odor 113 processing (Fig 2A) and running (Fig 2B) epochs. Data were taken from a session in which 114 animals performed at a high level (well-trained session). We observed that power in the 20-40 115 Hz and 4-8 Hz range observed during odor presentations showed significantly distinct patterns 116 across the proximodistal axis ( subjects ANOVAs: significant in 4 out of 5 subjects, see Table S1A). In contrast, 4-8 Hz power 120 was numerically higher in distal CA1 although the one-way ANOVA and linear trend analysis did 121 not reach significance, possibly due to an outlier (one-way ANOVA: F 3,12 = 1.0237, p = 0.4165; 122 Linear trend across electrodes: F 1,12 = 2.434, p = 0.1447; Individual subjects ANOVAs: 123 significant in 4 out of 5 subjects, with the fifth subject showing opposite pattern; see Table S1B). Hz bands observed during odor sampling were weak during running. Instead, we observed 128 strong oscillations in the 8-12 Hz theta range during running (consistent with numerous reports), 129 which increased toward proximal CA1 (one-way ANOVA: F 3,12 = 3.0296, p = 0.0711; Linear 130 trend across electrodes: F 1,12 = 8.944, p = 0.0113; Individual subjects ANOVAs: significant in 3 131 out of 5 subjects, see Table S2B). The running period was also characterized by increased 132 power in higher frequencies (> 24 Hz to avoid theta's first harmonic), which increased toward 133 distal CA1 (one-way ANOVA: F 3,12 = 3.7614, p = 0.0410; Linear trend across electrodes: F 1,12 = 134 10.21, p = 0.0077; Individual subjects ANOVAs: significant in 5 out of 5 subjects, see Table   135 S2A). Notably, the proximodistal pattern was significantly different between the two bands 136 (Electrode X band interaction: F 3,12 = 5.343, p = 0.0144). 137 20-40 Hz power increases with knowledge of the sequence 139 We then examined whether power in the 20-40 Hz range was linked with sequence memory 140 performance, and whether this association varied along the proximodistal axis. To do so, we 141 extended the previous analyses, which were applied to a well-trained session, to two  Table S3A).
153 154 We found that 20-40 Hz power increased with performance level (Fig 3C; one-way ANOVA: F 2,6 155 = 5.51, p = 0.0438; Individual subjects ANOVAs: significant in 2 out of 4 subjects; see Table   156 S3B), which complements our previous study showing learning-related differences in waveform 157 amplitude between InSeq and OutSeq trials (Allen et al., 2016). In addition, we found that this 158 performance effect did not significantly vary across the proximodistal axis, but instead scaled 159 with the local amplitude of the 20-40 Hz oscillation, suggesting this signal is present throughout 160 dorsal CA1 (data not shown). 161 162 20-40 Hz power varies with response type and accuracy 163 To shed light on the type of processing reflected by 20-40 Hz oscillations, we took advantage of 164 the four different trial types included in our paradigm: InSeq trials that were correctly or 165 incorrectly identified (InSeq+, InSeq-), and OutSeq trials correctly or incorrectly identified 166 (OutSeq+, OutSeq-). More specifically, we quantified power in the 20-40 Hz range (250ms 167 period before port withdrawal; averaged across the four electrodes) for each trial type 168 separately, as well as collapsing across trial types to test contrasts of particular interest (Fig 4). 169 To match previous plots, the data are first presented using a single value per animal to generate the group mean (Fig 4B), despite the increased variability induced by trial count discrepancies 171 across trial types (see Fig S1B and Table S4 for trial counts). To control for this, the data are 172 also presented using an approach in which we pooled trials from all animals and used a 173 sampling procedure to match trial count across trial types ( Fig 4C). InSeq responses than OutSeq responses (p = 0.002, permutation testing; Fig 4F). As expected, 194 the effect was in the same direction, but considerably more variable, when only considering a 195 single value per animal ( Fig 4D). Third, to test whether these oscillations are associated with 196 accurately performing the cognitive operations required on each trial, we compared power 197 between correct (InSeq+ and OutSeq+) and incorrect (InSeq-and OutSeq-) trials. Power was 198 significantly higher on correct trials (p = 0.002, permutation testing; Fig 4F). As above, the effect 199 was more variable but in the same direction when considering only one value per subject ( 100 Hz) bands during running epochs (most noticeably in distal CA1; Fig 2B). However, neither 236 slow (25-55 Hz) nor fast (60-100 Hz) gamma oscillations showed distinctive trial type-specific 237 information in the putative retrieval (250 ms prior to port entry) and encoding (110-300 ms 238 following port entry) windows, respectively ( Fig S2). However, we note that the slow gamma 239 range (25-55 Hz) overlaps with the observed 20-40 Hz oscillations prior to port withdrawal, 240 which may be a putative retrieval period. Overall, these findings suggest that the pattern of 241 distinct slow gamma and fast gamma oscillatory states observed in CA1 during spatial 242 navigation may be not be readily visible in the predicted epochs of a nonspatial sequence 243 processing task. However, it is possible that our experimental and analytical approach were not 244 optimal to directly test these effects.

246
In this paper, we examined oscillatory power in hippocampal region CA1 as rats performed a 247 complex sequence memory task to identify the oscillatory dynamics associated with nonspatial 248 information processing. The data presented here expand upon our previous report of 20-40 Hz power is higher for trials with an "in sequence" response (a presumed match between the 263 presented odor item and the predicted one) and during correct compared to incorrect trials. 264 Lastly, we do not find evidence that slow and fast gamma oscillations previously observed 265 during spatial exploration tasks are associated with specific trial types during the putative 266 encoding and retrieval epochs of the task, although we did not test other epochs for this effect. 267 We suggest that more work needs to be done to fully ascertain the role of slow and fast gamma 268 oscillations in nonspatial tasks. Altogether, these findings suggest that processing the temporal It is important to note that the nature of this complex experimental paradigm led to two potential 273 limitations to consider when interpreting the findings. First, the use of a nonspatial response 274 (hold/withdraw) prevented us from directly equating response duration across trial type. 275 Although the degree to which differences in response duration influence our findings cannot be 276 fully determined, we showed that response duration alone does not explain the differential 277 recruitment of 20-40 Hz power across trial types. Second, to ensure adequate performance, the task requires that the number of OutSeq trials be kept relatively low (otherwise it is unclear 279 which sequence is being tested). This results in an uneven number of observations across trial 280 types which could disproportionally influence a subset of our analyses. However, we controlled 281 for this possibility by conducting pooled analyses, which matched trial count across conditions 282 using a permutation sampling procedure, to ensure sufficient statistical power. Overall, we 283 believe these control analyses significantly mitigated the potential influence of these 284 confounding factors on the interpretation of our results.

374
Our group previously published using the same dataset, and a detailed description of the 375 methods can be found in Allen and colleagues (2016). The methods are summarized below. 376 Subjects. Five male Long-Evans rats were used in this study. Animals were water restricted for 377 optimum task engagement but were provided full access to water on weekends. Proper processing approaches were used to exclude data contaminated by electrical noise or artifacts. 397 As stated in the manuscript (page 16), 60Hz electrical noise was removed using a notch filter. 398 Trials with artifacts associated with bumping or touching the headstage (voltage values > 5 SD 399 above the mean) were automatically excluded. Note that this exclusion was performed before 400 (and blind to) analysis of the results.
Equipment. The apparatus used for this task consisted of a linear track with water ports on 402 either end for water reward delivery. One end of the maze contained an odor port (above the 403 water port) connected to an automated odor delivery system. Photobeam sensors detected 404 when the animal's nose entered and withdrew from the odor port, which respectively triggered 405 and terminated odor delivery. Separate tubing lines were used for each odor item, however, all  Odor sequence task. In this hippocampus-dependent task, rats were presented with series of 412 five odors delivered in the same odor port (Fig 1). In each session, the same sequence was 413 presented multiple times, with approximately half the presentations including all items "in 414 sequence" (InSeq; ABCDE) and the other half including one item "out of sequence" (OutSeq; 415 e.g., ABDDE). Each odor presentation was initiated by a nosepoke and rats were required to 416 correctly identify the odor as either InSeq (by holding their nosepoke response until the signal at 417 1.2 s) or OutSeq (by withdrawing their nosepoke before the signal; <1.2 s) to receive a water 418 reward. Animals were trained preoperatively on sequence ABCDE (lemon, rum, anise, vanilla, 419 and banana) until they reached asymptotic performance (>80% correct on both InSeq and 420 OutSeq trials; ~6 weeks). Following surgical recovery, electrophysiological data was collected 421 as animals performed the same sequence (ABCDE), followed by two consecutive sessions 422 using a novel sequence (VWXYZ; almond, cinnamon, coconut, peppermint, and strawberry). 423 Surgery. Rats received a preoperative injection of the analgesic buprenorphine (0.02 mg/kg, 424 s.c.) ~10 min before induction of anesthesia. General anesthesia was induced using isoflurane 425 (induction: 4%; maintenance: 1-2%) mixed with oxygen (800 ml/min). After being placed in the 426 stereotaxic apparatus, rats were administered glycopyrrolate (0.5 mg/ kg, s.c.) to help prevent 427 respiratory difficulties. A protective ophthalmic ointment was then applied to their eyes and their 428 scalp was locally anesthetized with marcaine (7.5 mg/ml, 0.5 ml, s.c.). Body temperature was 429 monitored and maintained throughout surgery and a Ringer's solution with 5% dextrose was 430 periodically administered to maintain hydration (total volume of 5 ml, s.c.). The skull was 431 exposed following a midline incision and adjustments were made to ensure the skull was level. 432 Six support screws (four titanium, two stainless steel) and a ground screw (stainless steel; positioned over the cerebellum) were anchored to the skull. A piece of skull ~3 mm in diameter 434 (centered on coordinates: -4.0 mm anteroposterior, 3.5 mm mediolateral) was removed over the 435 left hippocampus. Quickly after the dura was carefully removed, the base of the microdrive was 436 lowered onto the exposed cortex, the cavity was filled with Kwik-Sil (World Precision 437 Instruments), the ground wire was connected, and the microdrive was secured to the support 438 skull screws with dental cement. Each tetrode was then advanced ~900 m into the brain. 439 Finally, the incision was sutured and dressed with Neosporin and rats were returned to a clean 440 cage, where they were monitored until they awoke from anesthesia. One day following surgery, 441 rats were given an analgesic (flunixin, 2.5 mg/kg, s.c.) and Neosporin was reapplied to the 442 incision site. proximal and most distal, respectively) and two electrodes in between which were equidistant. 488 We confirmed the relative spatial distribution of these electrodes, as well as their localization were presented with series of five odors delivered in the same odor port (located at one end of a linear track). B. In 591 each session, the same sequence was presented multiple times, with approximately half the presentations including 592 all items "in sequence" (InSeq; ABCDE) and the other half including one item "out of sequence" (OutSeq; e.g., 593 ABDDE). Each odor presentation was initiated by a nosepoke and rats were required to correctly identify the odor as 594 either InSeq (by holding their nosepoke response until the signal at 1.2 s) or OutSeq (by withdrawing their nosepoke 595 before the signal; <1.2 s) to receive a water reward. After completion of each sequence (correctly or incorrectly), 596 animals were required to run to the other end of the linear track and return to the odor port before the next sequence 597 could be presented. C. Sample histology image showing the range of tetrode tip locations, which spanned much of 598 the proximodistal axis of CA1 (3 tip locations shown; red circles). For each animal, a set of four tetrodes equally 599 distributed across the proximodistal axis (with comparable locations across animals) was used for local field potential 600 activity analysis. 601 Figure 2. Odor sequence processing and running on a track are associated with distinct oscillatory states in 603 CA1, which vary across the proximodistal axis. A. Group peri-event spectrograms (n=5) during odor sampling 604 period (correct in sequence trials only) in four electrode locations along the CA1 proximodistal axis (0ms = port entry). 605 B. Group peri-event spectrograms from the same electrodes during the running period (0ms = center of the runway  subjects ANOVAs, respectively. See Table S3 for statistical results. preceding port entry ("retrieval" period) across all four trial types (InSeq+, OutSeq-, OutSeq+, and InSeq-) and 671 collapsing across correct (InSeq+ and OutSeq+) and incorrect (InSeq-and OutSeq-) responses. B. Fast gamma 672 power (60-100 Hz) for the 110-300 ms time window ("encoding" period) across the same trial types and correct vs 673 incorrect contrast. None of the group-level comparisons were significant (all p's < 0.05) and only one animal showed 674 a significant increase on incorrect trials, which was driven by high values on InSeq-trials (p < 0.0001). See Table S5  675 for permutation testing statistical report.