Heterogeneities in intrinsic excitability and frequency-dependent response properties of granule cells across the blades of the rat dentate gyrus

The dentate gyrus (DG), the input gate to the hippocampus proper, is anatomically segregated into three different sectors, namely the suprapyramidal blade, the crest region and the infrapyramidal blade. Although there are well-established differences between these sectors in terms of neuronal morphology, connectivity patterns and activity levels, differences in electrophysiological properties of granule cells within these sectors have remained unexplored. Here, employing somatic whole-cell patch-clamp recordings from the rat DG, we demonstrate that granule cells in these sectors manifest considerable heterogeneities in their intrinsic excitability, temporal summation, action potential characteristics and frequency-dependent response properties. Across sectors, these neurons showed positive temporal summation of their responses to inputs mimicking excitatory postsynaptic currents, and showed little to no sag in their voltage responses to pulse currents. Consistently, the impedance amplitude profile manifested low-pass characteristics and the impedance phase profile lacked positive phase values at all measured frequencies, voltages and for all sectors. Granule cells in all sectors exhibited class I excitability, with broadly linear firing rate profiles, and granule cells in the crest region fired significantly less action potentials compared to those in the infrapyramidal blade. Finally, we found weak pairwise correlations across the 18 different measurements obtained individually from each of the three sectors, providing evidence that these measurements are indeed reporting distinct aspects of neuronal physiology. Together, our analyses show that granule cells act as integrators of afferent information, and emphasize the need to account for the considerable physiological heterogeneities in assessing their roles in information encoding and processing.

excitability, temporal summation, action potential characteristics and frequency-dependent 48 response properties. Across sectors, these neurons showed positive temporal summation of their 49 responses to inputs mimicking excitatory postsynaptic currents, and showed little to no sag in 50 their voltage responses to pulse currents. Consistently, the impedance amplitude profile 51 manifested low-pass characteristics and the impedance phase profile lacked positive phase values 52 at all measured frequencies, voltages and for all sectors. Granule cells in all sectors exhibited 53 class I excitability, with broadly linear firing rate profiles, and granule cells in the crest region 54 fired significantly less action potentials compared to those in the infrapyramidal blade. Finally, 55 we found weak pairwise correlations across the 18 different measurements obtained individually 56 from each of the three sectors, providing evidence that these measurements are indeed reporting 57 distinct aspects of neuronal physiology. Together, our analyses show that granule cells act as 58 integrators of afferent information, and emphasize the need to account for the considerable 59 physiological heterogeneities in assessing their roles in information encoding and processing. 60

INTRODUCTION 70
The dentate gyrus (DG), the input gate to the mammalian hippocampus proper (Amaral et al. 71 2007;Andersen et al. 2006), has been implicated in spatial navigation, response decorrelation, 72 pattern separation and engram formation. Granule cells are the prominent neuronal subtype 73 within the DG, and have been studied extensively from the perspective of their intrinsic response 74 properties, plasticity profiles, in vivo response properties, their role as engram cells, their sparse 75 connectivity and sparse firing characteristics, and neurogenesis (Aimone et al. 2014;Amaral et 76 al. 2007;Bakker et al. 2008;Bliss and Lomo 1973;Danielson et al. 2017;Diamantaki et al. 77 2016;GoodSmith et al. 2017;Heigele et al. 2016;Kropff et al. 2015;Leutgeb et al. 2007;Li et 78 al. 2017;McHugh et al. 2007;Mishra and Narayanan 2019;Neunuebel and Knierim 2014;Sahay 79 et al. 2011;Senzai and Buzsaki 2017;Tonegawa et al. 2018). Electrophysiological recordings 80 from granule cells have characterized their response characteristics, including important 81 differences between mature and immature cell excitability (Fricke and Prince 1984;Krueppel et 82 al. 2011;Liu et al. 1996;Mody et al. 1992;Pedroni et al. 2014;Schmidt-Hieber et al. 2004;83 2007;Staley et al. 1992;van Praag et al. 2002). The dentate gyrus, within each location along its 84 dorso-ventral span, is anatomically segregated into three different sectors: the suprapyramidal 85 blade, the crest region and the infrapyramidal blade (Amaral et al. 2007). There are several well-86 established differences across these three sectors (Amaral et al. 2007), including morphological 87 differences (Claiborne et al. 1990;Desmond and Levy 1985;1982;Gallitano et al. 2016;Green 88 and Juraska 1985;Schneider et al. 2014), connectivity patterns (Claiborne et al. 1986), the ratio 89 of basket cells to granule cells (Seress and Pokorny 1981) and activity levels (Chawla et al. Granule cells across the three DG sectors exhibited class I excitability where they were 115 able to fire action potentials at arbitrarily low firing rates, with broadly linear profiles of firing 116 rate vs. current injection (f-I) curves. Together, the low-pass frequency-response characteristics, 117 the lack of positive impedance phase, and the linear f-I curve showing class I excitability point 118 to DG neurons across all these sectors acting as integrators of afferent information. We found no 119 significant differences in subthreshold response properties of these neurons across the three DG 120 sectors. However, we found that granule cells in the crest region fired less action potentials, in 121 response to suprathreshold current injections, when compared with their counterparts in the 122 infrapyramidal blade. Finally, we assessed correlations across the 18 different sub-and supra-123 threshold measurements for each of the three DG sectors, and found a large number of 124 measurement pairs showing weak pairwise correlations. This large subset of uncorrelated 125 measurements suggested that the set of measurements employed here in characterizing DG 126 granule cells are assessing distinct aspects of their physiology. Together, our analyses show that 127 dentate gyrus neurons act as integrators of afferent information, and emphasize the need to 128 account for the considerable heterogeneities inherent to this population of neurons in assessing 129 their physiology, including engram formation and their ability to perform channel and pattern 130 decorrelation. 131 132 133 134 135

Ethical approval 137
All experiments reported in this study were performed in strict adherence to the protocols cleared 138 by the Institute Animal Ethics Committee (IAEC) of the Indian Institute of Science, Bangalore. 139 Experimental procedures were similar to previously established protocols (Ashhad et al. 2015;140 Ashhad and Narayanan 2016;Das and Narayanan 2017;Narayanan et al. 2010;Narayanan and 141 Johnston 2008;2007;Rathour et al. 2016) and are detailed below. Animals were provided ad 142 libitum food and water and were housed with an automated 12 h light-12h dark cycle, with the 143 facility temperature maintained at 21 ± 2° C. All animals were obtained from the in-house 144 breeding setup at the central animal facility of the Indian Institute of Science. 145 146

Slice preparation for in-vitro patch clamp recording 147
Electrophysiological recordings (from a total of ~200 neurons) reported in this study were 148 obtained from 66 male Sprague-Dawley rats of 6-to 8-week age, with ~96% recordings from 149 rats in the 7-8 weeks age group. Rats were anesthetized by intraperitoneal injection of a 150 ketamine-xylazine mixture. After onset of deep anesthesia, assessed by cessation of toe-pinch 151 reflex, transcardial perfusion of ice-cold cutting solution was performed. The cutting solution 152 contained 2.5 mM KCl, 1.25 mM NaH2PO4, 25 mM NaHCO3, 0.5 mM CaCl2, 7 mM MgCl2, 7 153 mM dextrose, 3 mM sodium pyruvate, and 200 mM sucrose (pH 7.3, ~300 mOsm) saturated 154 with 95% O2 and 5% CO2. Thereafter, the brain was removed quickly and 350-μm thick near-155 horizontal slices were prepared from middle hippocampi (Bregma, -6.5 mm to -5.1 mm), using a 156 vibrating blade microtome (Leica Vibratome), while submerged in ice-cold cutting solution 157 saturated with 95% O2 and 5% CO2. The slices were then incubated for 10-15 mins at 34° C in a 158 chamber containing the holding solution (pH 7.3, ~300 mOsm) with the composition of: 125 mM 159 NaCl, 2.5 mM KCl, 1.25 mM NaH2PO4, 25 mM NaHCO3, 2 mM CaCl2, 2 mM MgCl2, 10 mM 160 dextrose, 3 mM sodium pyruvate saturated with 95% O2 and 5% CO2. Thereafter the slices were 161 kept in a holding chamber at room temperature for at least 45 min before the start of recordings. 162 A maximum of 6 middle hippocampal slices were obtained from each rat, and a maximum of 163 two neuronal recordings were obtained from each slice. For electrophysiological recordings, slices were transferred to the recording chamber and 167 continuously perfused with carbogenated artificial cerebrospinal fluid (ACSF/extracellular 168 recording solution) at a flow rate of 2-3 mL/min. All neuronal recordings were performed under 169 current-clamp configuration at physiological temperatures (32-35° C), achieved through an 170 inline heater that was part of a closed-loop temperature control system (Harvard Apparatus). The 171 carbogenated ACSF contained 125 mM NaCl, 3 mM KCl, 1.25 mM NaH2PO4, 25 mM NaHCO3, 172 2 mM CaCl2, 1 mM MgCl2, 10 mM dextrose (pH 7.3; ~300 mOsm). Slices were first visualized 173 under a 10× objective lens to locate the granule cell layer of the dentate gyrus and then 63× 174 water immersion objective lens was employed to perform patch-clamp recordings from DG 175 granule cells, through a Dodt contrast microscope (Carl Zeiss Axioexaminer). Whole-cell 176 current-clamp recordings were performed from visually identified dentate gyrus granule cell 177 somata, using Dagan BVC-700A amplifiers. 178 Borosilicate glass electrodes with electrode tip resistance between 2-6 MΩ (more often 179 electrodes with ~4 MΩ tip resistance were used) were pulled (P-97 Flaming/Brown micropipette 180 puller; Sutter) from thick glass capillaries (1.5 mm outer diameter and 0.86 mm inner diameter; 181 Sutter) and used for patch-clamp recordings. The pipette solution contained 120 mM K-182 gluconate, 20 mM KCl, 10 mM Hepes, 4 mM NaCl, 4 mM Mg-ATP, 0.3 mM Na-GTP, and 7 183 mM K2-phosphocreatine (pH 7.3 adjusted with KOH, osmolarity ~300 mOsm). Series resistance 184 was monitored and compensated online using the bridge-balance circuit of the amplifier. 185 Experiments were discarded only if the initial resting membrane potential was more depolarized  196 Narayanan et al. 2010;Narayanan and Johnston 2008;2007;Rathour et al. 2016). Input 197 resistance (Rin) was measured as the slope of a linear fit to the steady-state V-I plot obtained by 198 injecting subthreshold current pulses of amplitudes spanning -50 to +50 pA, in steps of 10 pA 199 (Fig. 1A). Owing to very high input resistance of many cells, and to avoid spike generation for 200 positive current injections, we also performed recordings in response -25 to +25 pA current 201 injection, in steps of 5 pA. To assess temporal summation, five0 α-excitatory postsynaptic 202 potentials (α-EPSPs) with 50 ms interval were evoked by current injections of the form I α Imax 203 t exp (-αt), with α 0.1 ms -1 (Fig. 1B). Temporal summation ratio (S α ) in this train of five 204 EPSPs was computed as Elast/Efirst, where Elast and Efirst were the amplitudes of last and first 205 EPSPs in the train, respectively. Percentage sag was measured from the voltage response of the 206 cell to a hyperpolarizing current pulse of 100 pA and was defined as 100 (1-Vss/Vpeak), where Vss 207 and Vpeak depicted the steady-state and peak voltage deflection from VRMP, respectively. 208 The chirp stimulus ( the Fourier transform of the Chirp15 stimulus formed the impedance amplitude profile (Fig, 1E). 214 The frequency at which the impedance amplitude reached its maximum was the resonance 215 frequency (fR). Resonance strength (Q) was measured as the ratio of the maximum impedance 216 amplitude to the impedance amplitude at 0.5 Hz (Hu et al. 2002). Total inductive phase (ΦL) was 217 defined as the area under the inductive part of the ZPP (Narayanan and Johnston 2008). A 100-218 pA hyperpolarizing current pulse was provided before the chirp current ( Fig. 1C) to estimate 219 input resistance ( ) and to observe and correct series resistance changes through the course of 220 the experiment. 221 To characterize these subthreshold physiological measurements (see Table 1) of granule 222 cells, recordings were performed at VRMP from the three major well-defined sectors of dentate 223 gyrus ( Fig. 2): suprapyramidal blade, crest region and infrapyramidal blade (Amaral et al. 2007). 224 The boundaries between these three regions were identified visually from the curvature of the 225 granule cell layer (Amaral et al. 2007). Specifically, cells recorded from the semi-circular region 226 between the two flat blades were assigned to the "Crest region", the flat blade closer to CA1 was 227 called the "Suprapyramidal Blade" and the one farther away from the CA1 was referred to as the 228 "Infrapyramidal Blade" (Fig. 2A). The recordings were uniformly distributed within the granule 229 cell layer, across deep, superficial and medial regions (along the hilus-molecular layer axis), of 230 these sectors. Images of cell location were stored for post-facto classification into one of three 231 granule-cell sectors. The characterization protocol to measure sub-threshold measurements was 232 repeated for a range of membrane voltages in a subset of cells, to assess the dependence of these 233 each of these pair-wise scatter plots and analyzed the distribution of correlation coefficients for 259 each population (Fig. 7). Qualitative descriptions about weak vs. strong correlations were 260 adopted from the definitions in the literature, with reference to the value of R (Evans 1996). 261 All data acquisition and analyses were performed using custom-written software in Igor 262 Pro (Wavemetrics), and statistical analyses were performed using the R computing package 263 (http://www.r-project.org/). Across figures, the statistics employed for data presentation was 264 consistent with the statistical test used to compare two populations of data. Specifically, when 265 data is reported as mean ± SEM, parametric tests (ANOVA followed by Tukey's HSD test, 266 Student's t test) were employed, and when data is reported as median (along with the entire 267 distribution of the data or the quartiles), we employed non-parametric tests (Kruskal Wallis, 268 Wilcoxon rank sum). To emphasize the heterogeneities, we have reported all data points, and not 269 just the statistics behind the data to avoid misinterpretations arising from reporting of summary In assessing the intrinsic response properties of DG granule cells across its different blades, we 278 performed patch-clamp electrophysiological recordings under current clamp mode at the cell 279 body of visually identified granule cells. We first characterized the response properties of these 280 neurons using a range of subthreshold electrophysiological measurements (Fig. 1). We measured 281 input resistance from the steady-state voltage response to pulse current injections of different 282 depolarizing and hyperpolarizing amplitude (Fig. 1A). Input resistance, a steady-state measure of 283 neuronal gain and excitability, falls inadequate in characterizing neuronal response properties to 284 ethologically relevant time-varying signals. Therefore, to assess neuronal response properties to 285 time-varying signals, we injected multiple current stimuli that mimicked excitatory postsynaptic 286 currents (EPSC) into neurons to understand temporal summation properties (Magee 1998;1999). 287 We found considerable temporal summation of EPSPs, with the value of temporal summation 288 ratio above unity (Fig. 1B). 289 As the dentate gyrus resides in an oscillatory neural network (Bland 1986;Buzsaki 2002;290 Colgin 2013;2016;Sainsbury and Bland 1981;Winson 1978;1974), we assessed neuronal 291 response properties to sinusoidal stimulus of different frequencies. A standard stimulus that is 292 employed in assessing frequency-dependent response properties is the chirp current stimulus 293 2002; Narayanan and Johnston 2008;2007;Pike et al. 2000). The gain of the system, measured 300 as impedance amplitude, reduced especially across lower frequencies with hyperpolarization of 301 membrane potential (Fig. 1D-E), and reflected in the maximal impedance amplitude |Z|max 302 reducing with hyperpolarization (Fig. 1E). 303 The advantage of employing impedance as an excitability measure is two-fold. First, 304 impedance amplitude measures neuronal excitability as a function of input frequency providing a 305 frequency-dependent excitability metric, and second, impedance phase provides the temporal 306 relationship between the voltage response and current input at various input frequencies (Mauro 307 1961;Mauro et al. 1970;Narayanan and Johnston 2008;Sabah and Leibovic 1969). We 308 computed the impedance phase at all measured frequencies, and found the voltage response to 309 lag the injected current at all membrane voltages (Fig. 1F). This is in striking contrast with CA1 which was zero at all voltages where chirp responses were recorded (Fig. 1F). 319 As functions of membrane voltage, both input resistance as well as maximal impedance 320 amplitude reduced with hyperpolarization, indicating a reduction in overall neuronal excitability 321 at hyperpolarized voltages (Fig. 1G). The resonance frequency (fR) was less than 1 Hz and the 322 resonance strength was close to unity at all measured voltages (Fig. 1H), indicating low-pass 323 response characteristics of this dentate granule cell. 324

325
Heterogeneities in subthreshold measurements across different blades of the dentate gyrus. 326 How do these different steady-state and frequency-dependent subthreshold measures of granule 327 cell response vary across the different blades of the dentate gyrus? To address this question, we 328 measured the different electrophysiological properties (Fig. 1) at resting membrane potential 329 from granule cells located within the three prominent sectors within the DG: the suprapyramidal 330 blade, the crest region and the infrapyramidal blade ( Fig. 2A). First, we found considerable cell-331 to-cell variability in each of these response properties, spanning all the three sectors ( Fig. 2B-I; 332 Table 1). For instance, whereas the median value of input resistance was around 150 MΩ across 333 all three sectors, the input resistance spanned a large range from tens to hundreds of MΩ (Fig.  334 2C). Second, we found that none of these subthreshold measurements were significantly different 335 across the three sectors ( Fig. 2B-I) implying the similarity in the degree of heterogeneity across 336 all sectors of DG. We quantified degrees of heterogeneity in each of these measurements 337 employing three statistical measures (standard deviation, interquartile distance and coefficient of 338 variation), and found them to be comparable across the three sectors (Table 1). 339 These measurements confirmed that DG granule cells across all three sub-structures lack 340 prominent sag (Fig. 2E) that is characteristic of the expression of resonance. This was consistent 341 with the resonance frequency of these neurons falling around 1 Hz (Fig. 2H), with the resonance 342 strength centered on unity (Fig. 2I). Together, these measurements indicate that DG granule cells 343 across all three sectors exhibited low-pass response characteristics. The temporal summation of 344 alpha current inputs showed the fifth EPSP to have higher amplitude than the first (temporal 345 summation ratio > 1) for most recorded neurons across all three sectors, indicating an enhanced 346 temporal summation in these neurons (Fig. 2F). However, cells within each sector exhibited 347 significant heterogeneity in terms of how they responded to the train of alpha current inputs, with 348 temporal summation ranging from a value just lower than unity to values greater than 1.5 in 349 certain cells (Fig. 2F). voltages? To address these questions, we altered the neuronal membrane potential employing DC 359 current injection, and recorded subthreshold measurements at five different voltage values (Fig.  360 3). We found neuronal excitability to reduce with increased hyperpolarization, inferred from 361 significant hyperpolarization-induced reductions in input resistance (Fig. 3A) and maximum 362 impedance amplitude (Fig. 3C) across all three sectors. These two measures of sub-threshold 363 excitability significantly reduced with hyperpolarization in membrane voltage (ANOVA 364 followed by Student's t test for each of the 10 unique pairs across the 5 voltages; p<0.001), but 365 were not significantly different across the three sectors (ANOVA; p>0.5). We noted that the 366 degree of heterogeneity in these two measurements was considerable across cells, even when 367 measured at a specific membrane voltage. Specifically, we quantified the degree of heterogeneity 368 using three statistical measures (standard deviation, interquartile distance and coefficient of 369 variation) associated with these two voltage-dependent excitability measurements (Table 2). For 370 each of the three sectors and at all the five voltages where measurements were performed, we 371 found the degree of heterogeneity in these measurements to be comparable to those observed at 372 VRMP (compare degree of heterogeneity in these measurements in Table 1 vs. Table 2). Together, 373 these results demonstrated the voltage-dependence of these two sub-threshold excitability 374 measurements, apart from providing evidence that the heterogeneity reported in Fig. 2 is not a 375 simple reflection of the heterogeneity in Vrest. 376 We noted that sag (Fig. 3B) and resonance frequency (Fig. 3D) did not change 377 significantly with membrane voltages, and resonance strength continued to center at unity (Fig.  378   3E). Statistically, none of sag, resonance frequency, resonance strength and total inductive phase 379 were significantly different across sectors or across voltages ( Fig. 3; ANOVA, p>0.5). 380 Furthermore, the total inductive phase was negligibly small across all voltages, confirming the 381 absence of an inductive phase lead in the impedance profile of granule cells across all three 382 sectors (Fig. 3F). Together these results demonstrated that DG granule cells exhibit low-pass 383 response properties with a distinct absence of inductive lead in the impedance phase, at all 384 subthreshold voltages and across the three sectors of the dentate gyrus. 385 386 Heterogeneities in firing properties and action potential measurements across different 387 blades of the dentate gyrus. 388 How do neuronal firing profiles and action potential properties vary across the different sectors 389 of the dentate gyrus? Are these suprathreshold measurements heterogeneous within each sector? 390 We injected depolarizing current pulses of different amplitudes to assess the firing profile (Fig.  391 4A-B) and several metrics associated with action potentials (Fig. 4C-D; Table 3) of DG granule 392 cells. Across the three DG sectors, the firing profile of dentate granule cells (Fig. 4B, Fig. 5B-F) 393 reflected class I excitability, where the neuron was capable of eliciting firing at arbitrarily low 394 frequencies (Hodgkin 1948;Ratte et al. 2013). In addition, beyond rheobase current (which was 395 between 50-150 pA in most recorded neurons; Fig. 5B-D), the firing rate increase as a function 396 of injected current was fairly linear (Fig. 5B). These observations, along with the low-pass 397 response characteristics and positive temporal summation observed earlier (Figs. 1-3), pointed to 398 the DG granule neurons acting as integrators of incoming information (Ratte et al. 2013). 399 We also observed considerable cell-to-cell variability in firing frequencies within each of 400 the three sectors. For instance, for a pulse current injection of 250 pA, the median firing rate of 401 these neurons was around 15 Hz across all three sectors (Fig. 5F). However, this rate varied over 402 a large range spanning 0 (no spikes) for certain cells to ~40 Hz in certain others. Thus, the mean 403 We next analyzed individual spikes from granule cells recorded from each of the three 412 sectors and derived metrics that quantified spike threshold, width, amplitude, depolarizing and 413 repolarizing kinetics. Similar to our observations with subthreshold properties (Fig. 2) and action 414 potential firing properties (Fig. 5), we found significant cell-to-cell variability in these 415 measurements even within a given sector (Fig. 6B-I). Across populations, we found action 416 potential amplitude to be significantly lower in granule cells from the crest region compared to 417 those from the two blades (Fig. 6C). Action potential threshold was significantly hyperpolarized 418 in crest region cells compared to those in the suprapyramidal blade (Fig. 6D). The latency to first 419 spike was significantly lower in the infrapyramidal population of granule cells compared to cells 420 in the crest region (Fig. 6F). Finally the peak derivative of the action potential trace was 421 significantly higher for cells in the crest region, compared to cells in the two blades (Fig. 6H). 422 We noted that although these differences were statistically significant, the range of values of 423 these measurements from neurons of all three sectors was not very different ( Fig. 6B-I). In 424 addition, as the differences in median (Fig. 6) or mean values (Table 3) of these measurements 425 across the sectors were not large (as a fraction of the respective range of each measurement), it 426 might be infeasible to infer large differences in action potential properties across DG sectors 427 based on these measurements (especially in light of heterogeneity in each measurement, 428 represented in Fig. 6 and quantified with different measures of degree of heterogeneity in Table  429 3). 430

431
A large proportion of sub-and supra-threshold measurements from all blades of the 432 dentate gyrus exhibited weak pairwise correlations 433 Are the different sub-and supra-threshold measurements correlated? Are these correlations 434 distinct across the different blades of the dentate gyrus? Correlations in measurements provide 435 clues about relationships between sub-and supra-threshold measurements, apart from pointing to 436 the possibility of similar ion channels mechanisms underlying these distinct measurements. We 437 plotted pairwise scatter plots of these intrinsic measurements from granule cells recorded from 438 the suprapyramidal blade (Fig. 7A), the crest region (Fig. 7B), the infrapyramidal blade (Fig. 7C) 439 and the pooled population containing cells from all sectors (Fig. 7D). We computed Pearson's 440 correlation coefficient for each of these pairwise scatter plots (Fig. 7A-D). Broadly, we found the 441 correlation matrices across the four populations to be fairly similar, with a large number of 442 measurement pairs showing weak pairwise correlations (between -0.4 to +0.4 (Evans 1996); see 443 inset histograms in Fig. 7A-D). A small set of pairs showed strong positive or negative 444 correlations, and this set was also broadly common across the 4 matrices ( Fig. 7A-D). Among  (Tables 1-3). Based on these metrics, we 466 noted that individual measurements had different degrees of variability (smallest for RMP and 467 largest for total inductive phase, based on coefficient of variation), but possessed similar degrees 468 of variability across the three sectors (Tables 1-3). Second, we performed principal component 469 analysis (PCA), a dimensionality reduction technique, on the dataset (shown in Fig. 7D) 470 comprised of all 18 measurements from the three sub regions (Fig. 8). We plotted the data from 471 the three sectors on the axes spanning the three dominant principal components (explaining 472 ~90% of the variance in the data). We found that data from these sectors did not form 473 independent clusters, pointing to similarity in heterogeneities of the data acquired from neurons 474 in the three sectors of the DG (Fig. 8). of repetitive action potential firing through constant current injection. Axons that were capable of 537 responding over a wide range of frequencies, especially at arbitrarily low frequencies were 538 designated as class I. Axons that were classified as class II exhibited a pronounced supernormal 539 phase, whereby the frequency of action potential firing was largely invariant to the injected 540 current amplitude after the first spike was elicited (for currents beyond the rheobase current). 541 Class III axons were those that elicited a second spike (beyond the first spike) only with 542 difficulty or not at all (Hodgkin 1948). This classification has provided an invaluable tool to 543 understand neuronal excitability, neural coding spike initiation dynamics, neuronal operating 544 characteristics and phase resetting curves in a broadly unified, with the ionic mechanisms 545 underlying these classes of excitability well understood (Das and Narayanan 2015;2014;2017;546 Das et al. 2017;Ermentrout 1996;Hodgkin 1948;Prescott et al. 2008a;Prescott et al. 2006;547 2008b;Ratte et al. 2013). Specifically, it is now recognized that the different classes of 548 excitability are consequent to cooperation or competition between fast inward currents and slow 549 outward currents. Cooperation between these two classes of currents yield class I excitability, 550 whereas competition yields class II/III excitability. Importantly, these classes of excitability have 551 been linked to the ability of neurons to act as integrators (class I) or as coincidence detectors 552 (class II/III), with the synergistic interactions among channels capable of sliding the operating 553 mode of a neuron along the integrator-coincidence detector continuum (Das and Narayanan 554 2015;2014;2017;Das et al. 2017;Ratte et al. 2013). 555 Our results demonstrate that the f-I curve of DG granule cells are capable of firing at 556 arbitrarily low frequencies, with firing frequency clearly dependent on input current injection 557 ( Fig. 5B), pointing to class I excitability characteristics. In addition, the absence of sag, 558 resonance and positive impedance phase also point to absence of a dominant slow outward 559 current that can contribute to class II/III excitability or coincidence detection capabilities (Das 560 and Narayanan 2015;2014;2017;Das et al. 2017;Krueppel et al. 2011;Ratte et al. 2013). 561 Together these results clearly point to dentate granule cell somata acting as integrators of afferent 562 information (also see (Aimone et al. 2011)). However, it should be noted that operating modes of 563 neurons could change in response to several factors, including activity-dependent plasticity of 564 channels and receptors, neuromodulation and changes in afferent activity patterns (Das and 565 Narayanan 2015;2014;2017;Das et al. 2017;Prescott et al. 2008a;Prescott et al. 2006;2008b;566 Ratte et al. 2013). Future studies should therefore assess the spike-triggered average of DG 567 neurons to understand their operating modes, the roles of different ion channels in regulating 568 operating mode across their somato-dendritic axis and the information processing strategies 569 employed by the DG neurons (Das and Narayanan 2015;2014;2017;Das et al. 2017;Krueppel 570 et al. 2011;Schmidt-Hieber et al. 2007). 571 Despite the well-established expression of the hyperpolarization-activated cyclic 572 nucleotide gated (HCN) channels in granule cells (Bender et al. 2003;Krueppel et al. 2011;573 Stegen et al. 2012;Surges et al. 2012), they don't express impedance resonance or positive 574 impedance phase (Figs. 1-3) unlike CA1 pyramidal neurons or entorhinal stellate neurons (Das 575 et al. 2017;Erchova et al. 2004;Hu et al. 2009;Hu et al. 2002;Mittal and Narayanan 2018;576 Narayanan and Johnston 2008;2007). Consistent with this, these neurons do not exhibit a strong 577 voltage sag in the response to pulse currents (Fig. 1A, Fig. 2E, Fig. 3B presented in our study also provide indirect evidence for the expression of HCN channels, 596 through the reduction of excitability at hyperpolarized voltages (Fig. 3), through a 597 hyperpolarization-induced suppression of gain that is dominant at low frequencies (Fig. 1E) and 598 through the reduction in the capacitive lag in the impedance profiles ( Fig. 1F) with 599 hyperpolarization. 600 601

Future directions 602
Future studies should explore all of somato-dendritic, infrapyramidal-suprapyramidal, dorso-603 ventral, and superficial-deep axes of the dentate gyrus, and characterize intrinsic heterogeneities 604 expressed not just in granule cells, but also in other cell types including the basket cells, the 605 mossy cells and the semilunar granule cells (Amaral et al. 2007;Williams et al. 2007). In 606 addition, these studies should explore if there are systematic gradients in voltage-gated and 607 ligand-gated channels across these different axes, which would alter the processing and encoding 608 strategies associated with these neurons. The assessment of these heterogeneities and gradients 609 are especially essential, given the gradients that are observed in intrinsic and synaptic properties