3.1 DMDC detects effects of age on dimension of biologically relevant band pass filtered hippocampal LFPs

We first applied DMDC to 2500 ms epochs of LFPs that were collected when subjects were at the center of the maze or in the outer portions of the arms. Prior to execution, signals were left unfiltered (No Filter) or band pass filtered for theta, slow gamma, fast gamma or SWRs. No main effects of age, drug, or location on Dimension or Volume were reported by DMDC for the unfiltered or theta filtered signal (Fig. 2; significant and non-significant statistics in Table 1). An age by drug interaction effect on Volume was observed for the theta band [F(1,124) = 3.4358, p = 0.0171]. Theta signals from Younger animals receiving CX614 exhibited decreased Volume compared to age match animals receiving vehicle (p < 0.05). Conversely, theta signals from Older animals receiving CX614 had increased Volume compared to age match animals receiving vehicle (p < 0.05).

Fig. 2

Dimension and Volume outcomes for biologically relevant oscillation frequency groups and the no filter control group separated by age and drug conditions. A Means and confidence intervals for Dimension across frequency groups. LFP signals from Older animals exhibited higher Dimension of slow gamma, fast gamma, and SWR filters compared to Younger animals. * represents a main effect of age. B Means and confidence intervals for Volume across frequency groups. Volume was unaffected by age, drug, or location conditions

Table 1 Summary of group effects for unfiltered, theta, slow gamma, fast gamma, and SWR signals. Statistically significant results are shown in bold red text. F is the ANOVA F ratio. P is the ANOVA probability value. NF = no filter. SG = slow gamma. FG = fast gamma. SWR = sharp wave ripple

Dimension for slow gamma was higher for Older animals compared to Younger animals [main effect of age: F(1,124) = 48.0816, p < 0.0001] with no effect of drug or location. There were no effects of age, drug, or location on slow gamma Volume. In the fast gamma band, signals from Older animals showed higher Dimension values than signals from Younger animals [main effect of age: F(1,124) = 4.7450, p = 0.0314] with no effect of drug or location. No effect of age, drug, or location was reported for fast gamma Volume. In the SWR band, signals from Younger animals showed higher Dimension values than signals from Older animals [main effect of age: F(1,124) = 5.6270, p = 0.0193] with no effect of drug or location. No effect of age, drug, or location was reported for fast gamma Volume. Thus, barring one interaction effect, Volume was not sensitive to age, drug, or location and Dimension was affected by age only and limited to the slow gamma, fast gamma bands, and SWR bands.

3.2 DMDC performed on control frequency bands produced effects that opposed outcomes from biologically relevant signal bands

Five control filters were applied to better understand the biological relevance of the initial bandpass analyses: 1) high pass filter run above 20 Hz (20 Hz HP) to observe the slow gamma, fast gamma, and SWR filters together (excluding theta), 2) low pass filter run below 100 Hz (100 Hz LP) to observe all of the previously studied spatial navigation filters together, 3) high pass filter was executed above 100 Hz (100 Hz HP) to exclude the spatial navigation bands previously studied, 4) bandpass from 100–135 Hz outside of the frequencies ascribed to spatial navigation bands to act as a band range control (similar to the band ranges for slow and fast gamma), and 5) bandpass from 4–100 Hz in conjunction with the SWR band pass (SWR + 4–100 Hz). Previous research has shown that there may be correlation in activity of SWR and the gamma ranges (Carr et al., 2012; Pfeiffer & Foster, 2015).

Effects of age, drug, and location on Dimension or Volume for these control conditions were sparse (Fig. 3, significant and non-significant statistics in Table 2). No main effects of age, drug, or location on Dimension or Volume were reported by DMDC for the 100 Hz LP or the 4–100 + SWR group. For the 20 Hz HP, and in an opposing direction to results from individually analyzed slow gamma, fast gamma, and SWRs, Younger animals had a significantly higher Dimension than Older animals [main effect of age: F(1,124) = 5.4401, p = 0.0214] with no effect of drug or location. An age x drug interaction effect on Dimension opposite to that shown for the theta band was observed for the 20 Hz high pass filter [F(1,124) = 4.8780, p = 0.0291]. Dimension was increased for signals from Younger animals receiving CX614 compared to those receiving vehicle (p < 0.05) and decreased for signals from Older animals receiving CX614 compared to those receiving vehicle (p < 0.05). Volume was not impacted by age, drug, or location for the 20 Hz HP filtered signals. For the 100 Hz HP group, signals from Younger animals had a significantly higher Dimensions than Older animals [F(1,124) = 111.1, p < 0.0001] with no effect of drug or location. No main effect of age, drug, or location on Volume was observed for the 100 Hz HP filtered signals. For the 100 to 135 Hz bandpass, Dimension was higher in signals captured from Younger animals than it was in signals from Older animals [ main effect of age: F(1,124) = 27.8362, p < 0.0001], with no effect of drug or location. There was no main effect of age, drug, or location on Volume for the 100 to 135 Hz bandpass filtered signals. Thus, similar to the DMDC output for biologically relevant oscillation bands, Volume was not sensitive to age, drug, or location for any of the control filtered signals and Dimension was affected almost exclusively by age (barring one age by drug interaction). Interestingly, main and interaction effects observed in the control conditions were all opposite of those observed in the biologically relevant signals.

Fig. 3

Dimension and Volume outcomes for control oscillation frequency groups separated by age and drug conditions. A Means and confidence intervals for Dimension across frequency groups. LFP signals from Younger animals exhibited higher Dimension of 20 Hz HP, 100 Hz HP, and 100–135 Hz filters compared to Older animals. # represents a main effect of location. B Means and confidence intervals for Volume across frequency groups. Volume was unaffected by age, drug, or location conditions. Signals recorded from Older animals contained more events in the theta band than signals recorded from Younger animals. * represents a main effect of age

Table 2 Summary of comparative statistics for Dimension and Volume values extracted from differently filtered control LFPs. Statistically significant results are shown in bold red text. F is the ANOVA F ratio. P is the ANOVA probability value3.3 More traditional power analyses revealed age and location effects for biologically relevant oscillation bands that contrasted DMDC outcomes

To compare DMDC output to more traditional power analyses, we calculated average power (Fig. 4A) and event rate (Fig. 4B) across biologically relevant oscillation frequency bands and compared across frequency ranges, age, drug, and location groups. The SWR band was not included because the wavelet calculation spans from 1 to 100 Hz to align with the analyses performed in McHail and Dumas (2020). Average power differed across frequency range groups [F(2,249) = 98.4353, p < 0.0001]. Average power was higher in theta than slow gamma (p < 0.05) or fast gamma (p < 0.05) and average power in slow gamma was higher than fast gamma (p < 0.05). This outcome was expected since the 1/f relationship of frequency and power in natural signals usually results in higher power at lower frequencies. There was a location effect for both slow [F(1,82) = 12.4241, p = 0.0007] and fast gamma bands [F(1,82) = 29.0083, p < 0.0001]. Slow gamma power was increased in animals in the center of the maze compared to animals in the outer portion of the maze (p < 0.05), while fast gamma power was decreased for animals in the center of the maze compared to animals in the outer portion of the maze (p < 0.05). There was no main effect of age, drug, or location on average power in the theta band. There were no interaction effects on average power seen amongst the three filters.

Fig. 4

Mean power and event rate outcomes for biologically relevant oscillation frequency groups separated by age and location conditions. A Means and confidence intervals for average power across frequency groups. Slow gamma power was greater in LFP signals recorded in the inner portion of the arms. Fast gamma power was greater in signals recorded in the outer portions of the maze arms. * represents a main effect of location. B Means and confidence intervals for event rate across frequency groups. * represents a significant main effect of age. # represents a significant main effect of location. ** represents significant post hoc results for filter type

There was a significant difference in event rate between the three frequency ranges [F(2, 249) = 158.6187, p < 0.0001] (Fig. 4B). The theta frequency range had more events than the slow gamma (p < 0.05) or fast gamma (p < 0.05) and slow gamma had more events than fast gamma (p < 0.05). Signals recorded from Older animals contained more events in the theta band [main effect of age: F(1,82) = 8.8462, p = 0.0039], but there was no effect of drug or location. There was also an interaction effect of age x drug on theta event rate [F(1,82) = 4.4891, p = 0.0374]. Signals from Older animals receiving CX614 contained more events in the theta band compared to animals receiving vehicle (p < 0.05), while Younger animals receiving CX614 had a lower event rate than vehicle counterparts (p < 0.05). There was no main effect of age, drug, or location on slow or fast gamma event rate. Thus, mean power and event rate results from more traditional analyses identify location but not age effects better than DMDC.

3.4 DMDC does not differentiate between alternation and non-alternation trials

We next determined if DMDC could distinguish between trials in which the animal subsequently alternated or did not alternate in its maze arm selection. Signals from outer arm regions were omitted and signals collected when the animal was facing the maze center were compared across age, drug, and alternation versus non-alternation categories for all filter types applied to the prior location comparisons. DMDC did not report any effect of alternation versus non-alternation on Dimension or Volume in the unfiltered, theta, slow gamma, fast gamma, SWR, 20 Hz High Pass, 100 Hz LP, 100 Hz HP, 100–135 Hz, or 4–100 Hz + SWR conditions (Table 3). Combined with the initial DMDC outcomes, it appears that DMDC better identifies more static or holistic features of the LFP signals (age) than short-term signal dynamics (location, alternation).

Table 3 Summary of comparative statistics for average power and event rate for biologically relevant LFP signals. Statistically significant results are shown in bold red text. F is the ANOVA F ratio. P is the ANOVA probability value. SG = slow gamma. FG = fast gamma3.5 DMDC reveals different dimensions and volumes for different LFP filters

Since different filtering types revealed differences in age and drug effects, we examined the contribution of filtering itself to the DMDC outcomes by directly comparing across filter types (unfiltered, theta, slow gamma, fast gamma, SWR, 20 Hz HP, 100 Hz LP, 100 Hz HP, 100–135 Hz, and 4–100 Hz + SWR) after collapsing across age, drug, location, and alternation variables. DMDC found multiple significant effects of filter type on Dimension [F(9,1250) = 1646.967, p < 0.0001] (Table 4). The only filter groups that did not differ were the no filter group compared to the fast gamma or 4–100 + SWR group. Also, the fast gamma group did not differ from SWR and the 100–135 Hz group did not differ from 4–100 + SWR.

Table 4 Tukey tests results comparing Dimension values across filters. * represents a p-value equal to or less than 0.05, ** represents a p-value less than 0.01, *** represents a p-value less than 0.0001, and n.s. represents any p-value greater than or equal to 0.05. NF = no filter. SG = slow gamma. FG = fast gamma. SWR = sharp wave ripple

DMDC also reported a main effect of filter type on Volume [F(9,1250) = 107.7878, p < 0.0001] (Table 5). Significant pairwise comparisons were more limited than for Dimension. Volume for the 20 Hz HP and SWR filter groups differed most frequently from the other filter groups (p < 0.0001 compared to all other filter groups but not each other). Volume for the slow gamma and fast gamma groups also differed from the 100 Hz HP group (each at p < 0.0001) (Fig. 5B) (Tables 4 and 5).

Table 5 Tukey tests results comparing Volume values across filters. * represents a p-value equal to or less than 0.05, ** represents a p-value less than 0.01, *** represents a p-value less than 0.0001, and n.s. represents any p-value greater than or equal to 0.05. NF = no filter. SG = slow gamma. FG = fast gamma. SWR = sharp wave rippleFig. 5

Dimension and Volume outcomes across oscillation frequency groups collapsed across age, drug, location, and alternation conditions. A Means and confidence intervals for Dimension across frequency groups. B Means and confidence intervals for Volume across frequency groups

When Volume was plotted against Dimension (Fig. 6), filters that included higher frequency ranges, like the 20 and 100 Hz high pass, the SWR band pass, appeared to have the highest Dimension while filters that included the lowest frequency ranges, such as theta, seemed to have the lowest Dimension (Fig. 6A).

Fig. 6

Dimension plotted against Volume when collapsed across age, drug, location, and alternation conditions. A Volume versus Dimension for biologically relevant frequency groups and the no filter control group. B Volume versus Dimension for control frequency groups

This trend for a relationship between frequency range and Dimension was apparent for Volume as well (Fig. 6B), though linear regressions for central tendency of the filter group [Dimension: R2 = 0.365, t(9) = 2.15, p = 0.641; Volume: R2 = 0.287, t(9) = 1.79, p = 0.1105] or variance of the filter group [Dimension: R2 = 0.296, t(9) = 1.83, p = 0.1042; Volume: R2 = 0.190, t(9) = 1.37, p = 0.2079] versus Dimension or Volume were not significant (Fig. 7).

Fig. 7

Regression calculation and 95% confidence limits for filter range versus mean Dimension or Volume or variability in Dimension or Volume. A Linear regression for Dimension Mean plotted against filter range. B Linear regression for Volume Mean plotted against filter range. C Linear regression for Dimension interquartile range (IQR) plotted against filter range. D Linear regression for Volume interquartile range (IQR) plotted against filter range