Layer-dependence of regular-spiker baseline ring rates and detection

With regard to the use of the terms 'spontaneous' or 'baseline' ring rate, it should be noted that in many studies, in particular for head-xed animals, spontaneous ring rates are reported at quiet rest, while in the present study the baseline ring included all time periods except for the times of interaction. Thus, the 'baseline' included sleeping, quiet wakefulness, grooming, running, free whisking, and touch of the oor and walls. Consequently, the baseline ring rates reported here should be taken as approximations. For a better comparability with previous studies, behavioral scoring during the complete recording sessions and subsequent restriction to the times of display of certain behaviors would be necessary.

Layer1 No L1 neurons could be recorded, and the number of MUs was also very low. The reasons for this are presumably both anatomical and methodological. First, removal of the dura might have damaged or temporarily disturbed supercial neurons. Second, directly after surgery, tetrodes were located where the rst spikes were recorded. This was done to already position the tetrode in the cortex and avoid supracortical bleeding or dura regrowth from damaging the tetrode during a later penetration. Third, apart from spontaneous ring, this spiking activity was determined as a response to whisker touch. At the same time, L1 neuronal density is low (Zhou and Hablitz, 1996), and L1 seems to be rather involved in top-down modulation of signals than bottom-up signal representation (Letzkus et al., 2011).

This would presumably lead to these neurons being particularly silent in anesthesia, while ketamine anesthesia through its inhibitory action on NMDA channels (Franks, 2008) already attenuates overall neuronal activity. Taken together, the method of tetrode positioning and the properties of L1 neurons presumably combined to render recording from L1 very rare.

At the same time, it can be speculated that top-down modulation plays a crucial role in generating the sex-specic responses reported here, and this might be partly mediated by L1.

If, alternatively, subtle motor dierences which were missed by the present analysis were the

as the inuence of M1 on BC computations, in particular by L5 cells, is known to be mediated by synapses formed in L1 (Xu et al., 2012). Thus, it would be interesting to record activity in L1 during social touch using other methods than tetrodes, in particular two-photon imaging.

Layer 2/3 Neurons in L2/3 are known to re at very low rates. Patch-clamp recordings have repeatedly shown mean ring rates of L2/3 neurons to be under 1 Hz. Thus, Brecht et al. (2003) report an average spontaneous ring rate of 0.07 Hz under urethane anesthesia, de Kock and Sakmann (2009) found an average rate of 0.31 Hz in quiet awake rats, and the average rate during quiet wakefulness in the studies of Poulet and Petersen (2008) and Crochet et al. (2011) using mice was 0.61 Hz and 0.2 Hz, respectively. This is in line with the ndings of Kerr et al. (2007), who, employing two-photon imaging in mice, showed L2/3 neurons to re with 0.25 Hz on average. Both patch-clamp and two-photon imaging do not have an inherent bias to sample highly active neurons, as has been proposed for extracellular electrode recordings. However, the neuronal ring rates found in L2/3 in this study put a supposed bias towards sampling an unusually active neuronal population, when recording with tetrodes, under question. For a population of 23 L2/3 RS, average baseline ring rate was 0.22 Hz, and thus in the same range as observed with intracellular methods. Although this does not preclude a bias, it indicates that in this study neurons with low ring rates could also be recorded, and that a bias, if present, was not very pronounced. This is supported by the example of a low-ring L5B RS which displayed very sparse interaction-related activity, as shown in Fig. 3.17B.

Layer 4 The activity of L4 neurons reported here was also surprisingly close to ring rates recorded intracellulary. The rst intracellular recordings in L4 of BC were performed under urethane anesthesia and indicated extremely low spontaneous ring rates of 0.05 Hz on average (Brecht and Sakmann, 2002). However, in a later study which compared ring in the awake and anesthetized states, mean ring rates in L4 were found to be 1.93 Hz in the awake state, and thus much higher than the 0.58 Hz observed during anesthesia (de Kock and Sakmann, 2009). These 1.93 Hz are very close to the 2.08 Hz mean ring rate observed here for L4 RS.

As has been the case for L2/3, this argues against a strong bias to detect high-ring neurons due to deployment of tetrode instead of intracellular recordings.

Layer 5 The mean spontaneous ring rates observed in this study were 2.77 Hz for L5A and 3.03 Hz for L5B. The nding that L5 units re at higher rates on average than units from other layers (see Fig. 3.30A), is supported by the literature. The precise ring rates reported, however, dier widely. In a patch-clamp study by Manns et al. (2004) in anesthetized rats, spontaneous ring rates were higher in L5B (mean 0.77 Hz) than in L5A (0.39 Hz). Stunningly, Jacob et al. (2012) also using patch-clamp recordings under anesthesia report average ring rates of 10.6 Hz for L5-RS and 11.8 Hz for intrinsic bursting cells (for details on the subtypes see 4.3.4). The rates reported by de Kock and Sakmann (2009) are in between these (1.62 Hz for slender-tufted and 4.12 Hz for thick-tufted pyramidal neurons), also recorded under

anesthesia. Thus, the responses observed here are within the range described by other authors using intracellular methods. In addition, as described in more detail in 4.3.4, the slightly higher ring rates observed in L5B than L5A are also in line with evidence on dierent anatomical and physiological cell types, and their distributions between the two sublaminae.

Layer 6 There is relatively little data available in the literature on the ring rates of BC L6 neurons. The baseline ring rate of 1.79 Hz observed here was higher than the 0.52 Hz described by de Kock and Sakmann (2009) in awake head-xed rats. However, the population of RS recorded in L6 in the present study was very low (n = 6), and thus an assessment of ring rate is dicult. Interestingly, in this layer more FS (n = 9) than RS were recorded. It is unclear, whether this is chance or reects the relative percentage and/or ring properties of L6 RS and FS. Apart from such possible dierences within the L6 neuronal population, the generally low number of SUs recorded in L6 might be due to two methodological factors.

First, recording quality decreased over the course of experiments, and thus in many cases it was dicult to record SUs in the end of experiments, when L6 had been reached. Second, as responses were particularly strong in L5B, there was an incentive to move slowly in this layer, correspondingly taking longer to reach L6.

General remarks The use of tetrodes has been repeatedly claimed to introduce a bias towards sampling of highly active neurons (Olshausen and Field, 2004; Chorev et al., 2009).

However, although this assumption seems reasonable and receives indirect support from many studies, there is little direct evidence. While studies supporting the extracellular recording bias do exist (Towe and Harding, 1970; Stone, 1973), they are methodologically in many ways incomparable to the tetrode/patching distinction in question. As pointed out before, many studies do support lower ring rates when neurons are recorded intracellularly, but others do not. This question can be answered through comparison of ring rates obtained using intra- and extracellular methods. Ideally, this would be tested on the same cells using both recording methods, but if this is not the case, the cells to be compared have to be minimally in the same brain area and layer, as well as of the same cell type. Barrel cortex would seem a uniquely well-suited area to settle this question, due to the rich literature accumulated on dierent aspects of its physiology, but the diversity of protocols does not permit a nal judgement on how strong this bias is in which layer, and in fact whether it exists at all.

These distinctions between protocols include awake vs. anesthetized animals⁷, with dierent anesthesia protocols further complicating the picture, freely-moving vs. head-xed animals, and dierent sampling over layers, which is not always reported (e.g. by Vijayan et al., 2010). In addition, both extra- and intracellular recording methodologies can dier, as do

7In an early work, Armstrong-James and George (1988b) described decreasing BC activity during anes-thesia. The strongly damping eects of anesthesia on BC activity were reproduced by Vijayan et al. (2010) using tetrode recordings during isourane anesthesia. They observed an average ring rate decrease from 6.1 Hz in the awake state to 0.15 Hz during anesthesia. While >75% of RS red with rates >1 Hz while awake, the same neurons always red at <1 Hz in anesthesia. To my knowledge, such a comparison is not available

e.g. single-contact tungsten microelectrodes and tetrodes. To name one example, Curtis and Kleinfeld (2009) recorded in rat BC with stereotrodes and observed a much higher average ring rate of 9 Hz. What is more, the authors hardly observed any units with a ring rate of <5 Hz. As they recorded in all cortical layers, the discrepancy in rates between this and the present study, as well as the very high number (>50%) of neurons with narrow spikes (see Fig. S6 in Curtis and Kleinfeld, 2009), is presumably indicative of discrepancies in the sampling properties of dierent extracellular recording methods and protocols.

As described above, neurons with very sparse interaction-related responses, as well as ring rates down to 0.05 Hz, were recorded. The comparison of the observed RS baseline ring rates with the literature not only conrms strong layer-specic dierences, but also indicates that absolute values are at least close to those obtained by intracellular recordings. This does, of course, not preclude bias completely, and in particular the practice of repositioning the tetrode if no spikes were detected could potentially have contributed to missing very silent and/or very sparsely responding neurons.