Optogenetic Mapping of Intracortical Circuits Originating from Semilunar Cells in the Piriform Cortex

Abstract

Despite its comparatively simple trilaminar architecture, the primary olfactory (piriform) cortex of mammals is capable of performing sophisticated sensory processing, an ability that is thought to depend critically on its extensive associational (intracortical) excitatory circuits. Here, we used a novel transgenic mouse model and optogenetics to measure the connectivity of associational circuits that originate in semilunar (SL) cells in layer 2a of the anterior piriform cortex (aPC). We generated a mouse line (48L) in which channelrhodopsin-2 (ChR) could be selectively expressed in a subset of SL cells. Light-evoked excitatory postsynaptic currents (EPSCs) could be evoked in superficial pyramidal cells (17.4% of n = 86 neurons) and deep pyramidal cells (33.3%, n = 9) in the aPC, but never in ChR⁻ SL cells (0%, n = 34). Thus, SL cells monosynaptically excite pyramidal cells, but not other SL cells. Light-evoked EPSCs were also selectively elicited in 3 classes of GABAergic interneurons in layer 3 of the aPC. Our results show that SL cells are specialized for providing feedforward excitation of specific classes of neurons in the aPC, confirming that SL cells comprise a functionally distinctive input layer.

Introduction

The complexity of the 6-layered mammalian neocortex has sparked interest in simpler cortical structures that might capture the essential information-processing capabilities of the cortex in a more accessible circuit (Shepherd 2011; Fournier et al. 2014). One of those simpler cortices is the primary olfactory (piriform) cortex, a 3-layered paleocortex that is involved in recognizing and remembering odors (Bekkers and Suzuki 2013). The piriform cortex (PC) has fewer types of neurons and, in general, less complex connectivity than does the neocortex, with an important exception. Whereas the neocortex is often arranged into repeating cortical modules, the PC lacks this kind of spatial order and instead expresses unusually profuse and wide-ranging associational (intracortical) connections (Haberly and Price 1978; Haberly and Bower 1984; Johnson et al. 2000; Davison and Ehlers 2011; Franks et al. 2011; Hagiwara et al. 2012; Luna and Morozov 2012). This diffuse intracortical connectivity has long been regarded as a basis for the computational power of the PC, which is thought to represent odors in some kind of sparse combinatorial code (Hasselmo and Barkai 1995; Hopfield 1995; Cleland and Linster 2009; de Almeida et al. 2013; McGinley and Westbrook 2013). For this reason, it is important to clarify the essential features of associational connections in the PC.

The densely packed principal cells in layer 2 of the PC are the main source of associational fibers, with additional contributions from the sparser principal cells in layer 3 (L3; Haberly and Price 1978; Haberly and Bower 1984; Protopapas and Bower 2000; Neville and Haberly 2004; Yang et al. 2004). Layer 2 principal cells are not homogeneous, but comprise at least 2 types, namely, semilunar (SL) cells, with their somata located in upper layer 2 (2a), and superficial pyramidal (SP) cells, in lower layer 2 (2b) (although there is likely to be a gradient of cell types between these 2 extremes; Suzuki and Bekkers 2006, 2011). Using extracellular electrical stimulation in slices, it has previously been found that SL cells tend to receive stronger afferent input from the olfactory bulb (OB) but weaker associational input from within the PC, whereas the converse is the case for SP cells (Suzuki and Bekkers 2006, 2011; Franks et al. 2011; McGinley and Westbrook 2011; Wiegand et al. 2011; Hagiwara et al. 2012). However, the contributions of SL and SP cells to the associational fiber system remain unclear, in part because the diffuse connectivity of these fibers makes it extremely difficult to find connections using direct electrophysiological methods (e.g., paired whole-cell recordings; Suzuki and Bekkers 2011). Here, we overcame these difficulties by using optogenetics.

We describe a novel transgenic mouse model, the 48L mouse, that allows us to selectively express channelrhodopsin-2 (ChR2) in a subset of SL cells. Using ChR2-assisted circuit mapping, we measure the strength of associational connections between these SL cells and a variety of postsynaptic target neurons in the anterior PC (aPC). We find that ChR2-expressing SL cells avoid targeting other SL cells, but provide strong monosynaptic associational excitation of SP cells and deep pyramidal (DP) cells in L3, as well as excitation of 3 classes of GABAergic interneurons in L3. These results clarify the intracortical connectivity of the aPC and introduce a new mouse model for the study of piriform circuits.

Materials and Methods

Animals

Experiments used a line of transgenic mice (48L) identified in a lentiviral enhancer trap screen (Kelsch et al. 2012; Shima et al. unpublished observations). The screen involved random insertion of a lentiviral vector containing a minimal heat shock promoter, tetracycline response transactivator (tTA), tet response element (TRE), mCitrine (mCit), and the woodchuck hepatitis post-transcriptional regulatory element flanked by long terminal repeats. The 48L line was initially identified from its expression of mCit⁺ neuronal somata in layer 2a of the aPC (Fig. 1A). Animals were genotyped at birth by PCR and by observing the expression of mCit within the eyes of positive animals. Animals used in experiments were bred on a C57BL6/J background by mating 48L-positive males with wild-type females.

Figure 1.

In 48L mice, cells labeled with mCit have their somata preferentially located in layer 2a of the aPC. (A) Confocal image projection of a coronal slice of the aPC of a 28-day postnatal (P28) 48L mouse showing mCit⁺ cells (green) concentrated in layer 2a. The lamina borders (dashed lines) were identified from a bright-field image of the same slice. LOT, lateral olfactory tract. (B) Single confocal section (optical thickness, 5 µm) of mCit in the aPC of a P24 48L mouse (left), Nissl stain of the same region (middle), and merge (right). Laminae are indicated at right. (C–E) Confocal image projections of coronal slices of the olfactory bulb (OB; C), hippocampus (HIP, D), and primary somatosensory cortex (S1, E) of a 48L mouse (P28). The OB (C) shows mCit⁺ cells in the glomerular (G) and granule cell (Gr) layers, but no mCit⁺ somata in the mitral cell (M) layer. The hippocampus (D) contains a few scattered mCit⁺ somata, but none appear in the granule cell layer of the dentate gyrus (Gr) or in the CA1 pyramidal cell layer (Py). The somatosensory cortex (E) contains mCit⁺ cells scattered across all layers. Some, but not all, of these cells have a clear pyramidal morphology (e.g., inset shows, expanded, the cell in the dashed box).

Some experiments used double transgenic mice that were bred by mating heterozygous 48L males with heterozygous GAD67-GFP (Δneo) females. GAD67-GFP (Δneo) mice express green fluorescent protein (GFP) specifically in neurons containing GABA (Tamamaki et al. 2003; Suzuki and Bekkers 2010b). Double transgenic (48L/GAD67-GFP) pups were identified at birth by green fluorescence both in the eyes and in the cerebellum, visible through the thin skull. The use of these animals allowed us to more easily target GFP-labeled interneurons located outside of layer 2a.

All animal housing, breeding, and surgical procedures were approved by the Animal Experimentation Ethics Committee and the Recombinant DNA Committee of the Australian National University and by relevant authorities at Brandeis University, and conform to the guidelines of the National Health and Medical Research Council of Australia and the National Institutes of Health.

Virus Injections

In most experiments, adeno-associated virus (AAV, serotype 2/9) was used to infect cells with a construct comprising TRE-driving ChR2 fused to mCherry. The TRE contained the Tet operator sequences fused to a minimal promoter. Newborn mCit⁺ 48L pups (P1–2) were anesthetized by hypothermia and immobilized on a custom-made clay mold. AAV-TRE-ChR2-mCherry (100–200 nL) was stereotaxically injected into the aPC in both hemispheres (coordinates from lambda: 1.95 mm lateral, 2.55 mm anterior, and 3.2 mm deep). Injections were done using nonfilamented borosilicate glass micropipettes (GC150, Harvard Apparatus; tip diameter 8–10 µm) attached to a 10-μL Hamilton syringe (Hamilton). In some experiments, an alternative viral construct using the ubiquitous cytomegalovirus (CMV) promoter to drive ChR2 expression (AAV-CMV-ChR2-mCherry) was injected into the OBs of P1–2 mCit⁺ 48L pups. Animals were allowed to recover for 20–27 days before use.

Brain Slice Preparation

All experiments used acute brain slices (300 µm thick) prepared from the aPC of wild-type C57BL6/J, control 48L, and virus-injected 48L or 48L/GAD67-GFP mice (P21–28). Standard methods of slice preparation were used (Suzuki and Bekkers 2006, 2011). Briefly, mice were deeply anesthetized with isoflurane (2% in oxygen), then rapidly decapitated. Coronal slices containing the lateral olfactory tract (LOT) were prepared using a vibrating slicer (Campden Instruments) under ice-cold cutting solution comprising (in mM) 125 NaCl, 3 KCl, 0.5 CaCl₂, 6 MgCl₂, 25 NaHCO₃, 1.25 NaH₂PO₄, 2 ascorbate, 3 pyruvate, and 10 glucose (osmolarity 305 mOs/kg), bubbled with 5% CO₂/95% O₂ (carbogen). The slices were incubated for 1 h at 34 °C in a holding chamber containing carbogen-bubbled artificial cerebrospinal fluid (ACSF; composition below), then were held at room temperature until required.

Electrophysiology

Visualized whole-cell patch-clamp recordings were made from neurons in the aPC using Dodt gradient-contrast videomicroscopy. The microscope (Zeiss Axioskop, Carl Zeiss) was also equipped with wide-field fluorescence to enable identification of fluorescently labeled neurons. Slices were continuously superfused at 2–3 mL/min with ACSF containing (in mM) 125 NaCl, 3 KCl, 2 CaCl₂, 1 MgCl₂, 25 NaHCO₃, 1.25 NaH₂PO₄, and 25 glucose (310 mOs/kg), bubbled with carbogen, and maintained at 32 ± 1 °C. The ACSF contained picrotoxin (100 µM) to block GABA_A receptor-mediated inhibitory postsynaptic responses. In many experiments, tetrodotoxin (TTX, 0.5 µM; Alomone), 4-aminopyridine (4-AP, 100 µM), or 6,7-dinitroquinoxaline-2,3-dione (DNQX, 10 µM; Tocris) were also added to the ACSF. Patch electrodes had resistances of 4–6 MΩ when filled with internal solution containing (mM): 135 KMeSO₄, 7 NaCl, 0.1 EGTA, 2 Na₂ATP, 2 MgCl₂, 0.3 Na₃GTP, and 10 HEPES at pH 7.2, supplemented by 0.4% biocytin (295–300 mOs/kg). Unless stated otherwise, all compounds were obtained from Sigma-Aldrich.

Data were acquired using a Multiclamp 700B amplifier (Molecular Devices). For current-clamp recordings, the cell was allowed to remain at its resting membrane potential. Bridge balance and capacitance neutralization were carefully adjusted and checked for stability. A series of current steps (duration 500 ms, amplitudes ranging from −80 to 500–2000 pA in increments of 10–40 pA) was applied to identify neuronal firing patterns. For voltage-clamp recordings, the soma was held at −70 mV. Series resistance was monitored for stability, but series resistance compensation was not used. Voltage or current traces were filtered at 10 kHz and digitized at 20–50 kHz by an ITC-18 interface (Instrutech) under the control of Axograph (Axograph Scientific).

Optical Stimulation

Wide-field stimulation of ChR2 with blue light (∼490 nm) was achieved with a mercury lamp gated by a shutter (Uniblitz, Vincent Associates) and directed through a GFP filter set, which produced approximately 8 mW/cm² at the slice. Measurements with a photodiode showed that the rise time of the shutter-gated light flash was 2.1 ± 0.02 ms (mean 20–80% rise time, n = 10), necessitating a relatively long light flash (halfwidth 8–10 ms) to achieve a light response that reached a plateau. Control experiments in which the duration of the command pulse to the shutter was systematically varied while recording light-evoked currents in a ChR2-positive cell confirmed that this shutter duration was the minimum that elicited a maximal response amplitude. The relatively slow shutter speed was not a concern for the circuit mapping experiments reported here because connectivity was defined by the time-integrated response (i.e., charge; Fig. 5C). The illuminated region was circular with a diameter of approximately 400 µm. Typically, 20–50 flashes were applied at 10 s intervals while recording the light-evoked excitatory postsynaptic current (EPSC) in a ChR2-negative cell in voltage-clamp mode. Control experiments using pairs of flashes at different interflash intervals confirmed that light-evoked responses were fully recovered when evoked at 10 s intervals (not illustrated).

Figure 5.

Measurement of light-evoked EPSCs shows that SP and DP cells receive monosynaptic excitatory input from ChR2-expressing SL cells, but SL cells receive no input of this kind. (A) Cartoon summarizing the recording configurations and the main findings. Whole-cell voltage-clamp recordings were made from mCit⁻ SL, SP, and DP cells while flashing blue light over all layers of the aPC (represented by gray circles). SP and DP cells are the main recipients of inputs from ChR2-expressing SL cells (axon and boutons indicated by gray lines and triangles). (B) Typical raw data trace from an SP cell in which an EPSC was recorded in response to a flash of light (gray bar). Each trace was analyzed by integrating over two 150-ms long windows, one starting at the light flash (“EPSC”) and the other starting 250 ms later (“Backgnd”). (C) Frequency histogram of the average charge measured in the “EPSC” window for each of the 129 cells in the dataset (gray histogram). Superimposed on this is the frequency histogram of average charge measured in the “Backgnd” window (black outline), scaled to the same peak as the gray histogram. Vertical dashed line is the −70 pC detection threshold for identifying a light-evoked EPSC (see Materials and Methods). Inset shows a frequency histogram of the same “Backgnd” window data as in the main panel, calculated using a smaller bin width (12 cf. 20 pC in the main panel). The superimposed gray curve is a fit of the extreme value (Gumbel) distribution with µ = 0.073 and σ = 15.15. Vertical dashed line indicates the −70 pC detection threshold. (D) Averaged light-evoked responses (black traces), each superimposed on 2 typical single episodes (gray traces), measured in SL, SP, and DP cells (rows), showing 3 examples for each cell type (columns). Each black trace is an average of 20–50 responses to light (small gray bars at top), recorded in the presence of 0.5 µM TTX, 100 µM 4-AP, and 100 µM picrotoxin. (E) Bar plot summarizing the results of all experiments like in D. Numbers above bars are: number of that cell type exhibiting a light-evoked EPSC in TTX + 4-AP/total number of that cell type tested. *P < 0.02, χ² 2 × 2 contingency test.

Narrow-field stimulation of specific layers of the aPC (Fig. 6) was achieved with a digital projector (Epson EB-1770W) controlled by custom Matlab code (Mathworks). This method produced up to 6 mW/cm² at the slice with a 20–80% rise time of 12.4 ± 0.1 ms (n = 8). A photodiode placed in the light path was used for synchronization of the electrophysiological recording to the timing of the light flash.

Figure 6.

Layer-specific light stimulation of ChR2⁺ inputs onto SP cells is consistent with the idea that SL cells provide associational synaptic connections in layers 1b, 2, and 3. (A) Cartoon summarizing the experimental design and main finding. Whole-cell voltage-clamp recordings were made from SP cells while applying flashes of blue light to specific layers of the aPC (light gray shading). Excitatory inputs from ChR2⁺ SL cells (gray lines and triangles) are confined to layers 1b, 2, and 3. (B) Typical averaged traces recorded in an SL cell while flashing blue light in a band restricted to each layer (1a, 1b, 2, or 3, indicated at left). Each trace is an average of 20 episodes. Recordings were made before and after perfusing the bath with 0.5 µM TTX and 100 µM 4-AP (black and gray traces, respectively). The light flash (provided by a digital projector) was approximately 30 ms in duration (horizontal gray bar, top). (C) Mean light-evoked EPSC charge averaged across all responsive SP cells in the dataset (n = 9) plotted against the layer to which the light was applied. Black and gray plots summarize measurements in the absence and presence, respectively, of TTX + 4-AP. Error bars show SEM. *P < 0.05 compared with response when light is applied to layer 1a.

Histology

At the conclusion of the recording, the patch electrode was carefully retracted while maintaining the seal. The slice was fixed for 1 h in 4% paraformaldehyde in phosphate buffer, then stored in phosphate-buffered saline at 4 °C until processing. Neuronal morphology was revealed using either an ABC kit (Vector Laboratories) with diaminobenzidine, or streptavidin labeled with Alexa Fluor 594 (Life Technologies). Cell tracing was done manually using the Neurolucida tracing system (MBF Bioscience), which was also used for morphological analysis.

Confocal fluorescence imaging was done using a Zeiss Pascal or Nikon A1 microscope with a ×20/0.75 NA or ×10/0.45 NA objective. The mCit detection used 488 nm excitation and a 505–530 nm emission filter, whereas mCherry and Alexa Fluor 594 detection used 543 nm excitation/560–615 emission or 561 nm excitation/595–645 nm emission, respectively. Image stacks were acquired at 10 µm intervals through the slice, then z-projections were calculated for illustration (Figs 1 and 4). The efficiency of ChR2 expression was estimated by counting the fraction of mCit⁺ cells that were also mCherry-positive within the slice showing the highest expression of mCherry.

Figure 4.

Functional ChR2 can be selectively expressed in mCit⁺ SL cells. (A) Confocal image projection of a coronal slice of the aPC of a 48L mouse (P28) that had been injected with AAV-TRE-ChR2-mCherry at P2, showing tTA-mCit⁺ cells (green, top left), ChR2-mCherry⁺ cells (red, top right), and merge (bottom left). The expression of ChR2-mCherry is mostly restricted to the mCit⁺ SL cells. (B) Whole-cell voltage-clamp recordings of photocurrents elicited by flashes of blue light of increasing duration (10–500 ms) in a ChR2-mCherry⁺ SL cell. Bath solution contained TTX to block Na channels.

Nissl processing used NeuroTrace 530/615 red fluorescent Nissl stain (Life Technologies, 1 : 300 for 20 min). A single optical section (5 µm thick) was acquired using the Nikon microscope. The number of mCit⁺ cells within a region of interest in the superficial one-third of layer 2 was counted, and this number was divided by the number of Nissl⁺ cells in the same region to obtain the fraction of mCit⁺ cells. Details of this method, including detection criteria and the treatment of boundary effects, have previously been described (Suzuki and Bekkers 2010b).

Analysis of Electrophysiological Data

All electrophysiological analysis was done using Axograph, Matlab, or Igor Pro. Input resistance was calculated from the voltage responses to a series of hyperpolarizing current steps in current-clamp mode. The membrane time constant was measured by fitting a single exponential to the voltage response to an 80-pA hyperpolarizing current step applied at the resting potential. The action potential (AP) burst index was calculated as Δt₇/Δt₁, where Δt₁ is the time interval between the first and second APs, and Δt₇ is the interval between the seventh and eighth APs, using the first episode above rheobase that contained at least 8 APs.

Light-evoked responses were analyzed as follows. First, each EPSC episode in a trial (typically 20–50 episodes per trial) was digitally filtered at 2 kHz, then adjusted to a 5-ms baseline immediately before light onset and integrated over a 150-ms-long time window starting at light onset to obtain the light-evoked charge response (Fig. 5B). These single-episode values were then averaged together to obtain a mean EPSC charge for that trial. Similar measurements were made 250 ms after light onset to obtain the background response due to ongoing activity in the absence of light stimulation, for example, miniature EPSCs (Fig. 5B). The sets of background responses for each cell type (SL, n = 34; SP, n = 86; DP, n = 9) were not significantly different from each other (Kruskal–Wallis test, P > 0.05); hence, for initial analysis, the background responses for all cell types were merged. A histogram of these values showed a skew toward increasing negative values, reflecting the inward currents due to miniature EPSCs (Fig. 5C, inset). To choose a suitable detection threshold for light-evoked EPSCs, we first fitted the distribution of background activity to both a gamma distribution with a constant offset and an extreme value (Gumbel) distribution. (Fig. 5C, inset, shows the fit of the Gumbel distribution.) Both fits gave very similar results in the analysis described below. Based on these fits, the upper one-percentile (P = 0.01) and upper 0.1-percentile (P = 0.001) thresholds corresponded to absolute detection thresholds of −70 and −105 pC, respectively. The −70 pC threshold is shown in Figure 5C (vertical dashed line, main panel, and inset). In Figure 5C (main panel), the histogram of background activity (black) is superimposed on a histogram of the mean light-evoked EPSC charges (gray), merged across all cells, after scaling the peak of the former to equal that of the latter. The histogram of light-evoked charge shows a clear separation between the null light-evoked responses, clustered around the origin, and the long tail of cells for which a light-evoked EPSC was elicited. Visual inspection confirms that the choice of a detection threshold of either −70 pC (vertical dashed line) or −105 pC gives a good separation between background and response. We also tested the effect of treating the background for each cell type separately, rather than merging them into a single distribution. Fits of the gamma distribution were used to estimate detection thresholds as before. We found that the conclusion for SL cells was unchanged, whereas the estimated number of light-responsive SP cells was increased by 7 and that of DP cells was decreased by 1. Finally, for each cell with a mean response greater than the threshold, the peak amplitude of the light-evoked EPSC was measured by averaging the mean current trace over a 1-ms-long window around the peak.

Average results are given as mean ± SEM (n = number of cells, slices, or animals). Statistical comparisons used the unpaired two-tailed Student's t-test, Kruskal–Wallis test, Kolmogorov–Smirnov test, χ² 2 × 2 contingency test, or one-way analysis of variance (ANOVA) with Tukey's test for multiple comparisons, all with P-values as indicated.

Results

Expression of mCit⁺ Neurons in the 48L Mouse Line

An enhancer trap screen of mouse lines with random lentiviral insertions of reporter constructs revealed one strain, 48L, that expressed the mCit reporter in a defined population of cells that had their somata mostly restricted to the superficial half of layer 2 (layer 2a) of the aPC (Fig. 1A). It has previously been reported that a class of glutamatergic neurons, SL cells, are concentrated in layer 2a of the aPC (Suzuki and Bekkers 2006, 2011). We therefore wondered if the 48L mouse serendipitously expressed tTA-mCit only in SL cells.

In support of this idea, we observed what appeared to be dendritic labeling confined to layer 1 of the aPC of 48L mouse (Fig. 1A). This expression pattern is consistent with the known morphology of SL cells, which typically have few basal dendrites and extend apical dendrites into layer 1 (Haberly 1983; Haberly and Bower 1984; Neville and Haberly 2004). Another feature was a prominent band of mCit fluorescence in layer 1b (Figs 1A,B and 4A), which was probably due to labeled axons from mCit⁺ cells forming intracortical (associational) connections within the aPC (Neville and Haberly 2004; Suzuki and Bekkers 2011; see also next section).

The fraction of SL cells labeled with mCit was estimated from cell counts in Nissl-stained coronal sections of the aPC (Fig. 1B). Focusing on the upper one-third of layer 2, which is where “classic” SL cells are concentrated (Suzuki and Bekkers 2011) and where most mCit⁺ cells were located (Fig. 1B), mCit⁺ cells comprised 46 ± 2% of Nissl-labeled cells (range 36–64%, n = 16 slices from 2 animals aged P23 and P24). A few mCit⁺ cells were also present in layer 1b of the aPC (Fig. 1A). The fraction of mCit⁺ cells declined in older animals (>P30) and labeled cells were largely absent in 48L mice older than about P50 (data not shown).

Scattered cells expressing mCit were also seen in the OB of 48L mice, mainly in the granule cell layer and around the glomeruli (Fig. 1C). However, mitral/tufted cells, which provide input to the PC via the LOT, were never labeled. This finding is consistent with the absence of mCit label in the LOT in the aPC (Fig. 1A). Cells expressing mCit were also seen at low density in the hippocampus, but were not present in the granule cell layer of the dentate gyrus or in the pyramidal cell layer of regions CA1–CA3 (Fig. 1D). Neocortical areas contained mCit⁺ cells scattered across all layers (e.g., primary somatosensory cortex, Fig. 1E). Some of these cells had a clear pyramidal morphology (e.g., Fig. 1E, inset), but we did not confirm their identity by other means.

Cell Tracing Confirms that mCit Labels a Subpopulation of SL Cells in the 48L Mouse

We next used visualized whole-cell patch clamping in slices to load biocytin into mCit⁺ and mCit⁻ cells in layer 2 of the aPC of 48L mice, allowing recovery of their dendritic and axonal morphologies (Fig. 2). Recording first from neurons with their somata located in the most superficial one-third of layer 2 (where SL cells are concentrated in wild-type mice; Suzuki and Bekkers 2011), we found that both mCit⁺ and mCit⁻ cells had the profuse apical dendrites and sparse basal dendrites that are typical of SL cells (Fig. 2A, left and middle panels; dendrites shown in black). In contrast, mCit⁻ cells with their somata in deep layer 2 (2b) had the bitufted dendritic morphology that is typical of SP cells (Fig. 2A, right panel).

Figure 2.

mCit⁺ cells in superficial layer 2 of the aPC have a dendritic morphology that is typical of SL cells. (A) Dendrites (black) and axons (gray) of typical mCit⁺ cell (left) and mCit⁻ cell (middle) in upper layer 2 of the aPC of a 48L mouse. Note the presence of 2 simple basal dendrites in the mCit⁻ cell. At right is the dendritic morphology of a typical SP cell in deep layer 2. The lamina labels at right apply to all neurons shown in this panel. (B) Averaged polar plot of the dendrites (solid lines) and axons (dashed lines) of mCit⁺ (gray) and mCit⁻ (black) cells in superficial layer 2. Each line is an average of 8–10 cells. (C) Averaged total lengths of apical dendrites, basal dendrites, and axons of mCit⁺ (gray) and mCit⁻ (black) cells in superficial layer 2. Each bar is an average of 8–10 cells. *P < 0.01. (D) Top, Dendrites of a typical SL cell without (left) and with (right) basal dendrites, both from the aPC of a wild-type mouse. Bottom, Averaged total lengths of apical and basal dendrites, measured in SL cells without (left, n = 8 cells) and with (right, n = 9) basal dendrites. NS, not significantly different (P = 0.9) compared with mean apical dendrite length for mCit⁺ cells in C; *P = 0.03 compared with mean apical dendrite length for mCit⁻ SL cells in C. The y-axis in C also applies to D.

We did, however, notice 2 minor differences between mCit⁺ and mCit⁻ SL cells. First, mCit⁺ cells always lacked basal dendrites, whereas mCit⁻ cells had 1–3 simple basal dendrites (compare left and middle panels, Fig. 2A). Second, the total length of apical dendrites was greater in mCit⁻ cells than in mCit⁺ cells. These 2 differences were quantified in an averaged polar plot (Fig. 2B) and a bar plot (Fig. 2C). The polar plot shows that mCit⁻ cells have a small lobe corresponding to the basal dendrites (Fig. 2B, solid black line, lobe oriented at 270°), whereas mCit⁺ cells lack this lobe (Fig. 2B, solid gray line; each plot an average of n = 10 cells). The polar plot also shows the larger extent of apical dendrites in mCit⁻ cells (Fig. 2B, black lobe oriented at 90° is larger than the gray lobe averaged from mCit⁺ cells). The bar plot (Fig. 2C) confirms that the total lengths of apical and basal dendrites were significantly larger in layer 2a mCit⁻ cells than in mCit⁺ cells (apical: mCit⁻, 2962 ± 167 µm; mCit⁺, 2311 ± 134 µm, P = 0.007; basal: mCit⁻, 558 ± 152 µm; mCit⁺, 0 ± 0 µm, P = 0.005; all n = 10 cells, unpaired two-tailed t-test). Thus, the presence of mCit appears to be associated with slightly reduced dendritic outgrowth, both apical and basal.

Is mCit expression also associated with altered outgrowth of axons? This question was addressed by tracing the axons of the biocytin-filled mCit⁺ and mCit⁻ cells analyzed above (while keeping in mind that axon lengths are likely to be underestimated in slices because of the amputation of long collaterals; Fig. 2A, left and middle panels; axons shown in gray). As suggested by the polar plot (Fig. 2B, dashed lines) and confirmed in the bar plot (Fig. 2C), the total axonal length was not significantly different between mCit⁺ and mCit⁻ SL cells (mCit⁺: 5427 ± 549 µm, n = 8 cells; mCit⁻: 5314 ± 851 µm, n = 9 cells; P = 0.9, unpaired two-tailed t-test). Note that the presence of mCit⁺ axon collaterals in layer 1b (e.g., Fig. 2A, left) is consistent with the bright band of fluorescence visible in this layer (Figs 1A and 4A).

Is the occurrence of 2 variants of SL cells (with and without basal dendrites) an artifact of the 48L mouse? This question was addressed by tracing the dendrites of SL cells in slices of the aPC from wild-type mice (Fig. 2D). Neurons chosen for dye-filling were selected randomly, but in such a way that their somata were distributed uniformly across the most superficial one-third of layer 2, that is, the same territory from which we drew the mCit⁺ and mCit⁻ cells discussed above (distributions of soma locations were not significantly different: P = 0.45, n = 17 neurons in wild-type slices, n = 20 neurons in 48L slices, Kolmogorov–Smirnov test). We found that approximately half of the wild-type SL cells lacked basal dendrites (n = 8 cells out of 17, e.g., Fig. 2D, left), whereas the remainder had 1–3 simple basal dendrites (n = 9 cells out of 17, e.g., Fig. 2D, right). This proportion is similar to the fraction of basal dendrite-lacking mCit⁺ cells in the upper one-third of layer 2 from 48L slices (46%; previous section).

These 2 variants of wild-type SL cells seem broadly similar to the mCit⁺ and mCit⁻ SL cells observed in 48L tissue. For example, the SL cells lacking basal dendrites had apical dendrites of similar total length (wild-type apical: 2340 ± 102 µm, n = 8 cells; mCit⁺ apical: 2311 ± 134 µm, n = 10 cells; P = 0.9; unpaired two-tailed t-test; Fig. 2D, left). On the other hand, the SL cells with basal dendrites differed somewhat, with the wild-type SL cells having slightly shorter dendrites on average than the mCit⁻ SL cells, although the significance was marginal (wild-type apical: 2408 ± 154 µm, n = 9 cells; mCit⁻ apical: 2962 ± 167 µm, n = 10 cells; P = 0.03; wild-type basal: 230 ± 41 µm, n = 9 cells; mCit⁻ basal: 558 ± 152 µm, n = 10 cells; P = 0.06; Fig. 2D, right).

In summary, it seems unlikely that the presence of the lentiviral reporter construct in mCit⁺ SL cells in 48L mice causes the observed loss of simple basal dendrites, because a similar pattern of dendritic morphologies is also observed in wild-type SL cells. It is more likely that mCit expression is a consequence, rather than a cause, of a genetic program that drives morphological diversity in the aPC.

Similar Electrical Properties of mCit⁺ and mCit⁻ SL Cells

We next asked whether the morphologically identified SL cells described in the previous section also had the electrophysiological hallmarks of SL cells. Whole-cell current-clamp recordings were made from the 4 variants of putative SL cells identified above: mCit⁺ and mCit⁻ cells located in superficial layer 2 in 48L mice, and cells with or without basal dendrites located in superficial layer 2 in wild-type mice (same dataset as in the previous section; example recordings in Fig. 3A).

Figure 3.

Putative SL cells in 48L mice have intrinsic electrical properties that are typical of SL cells and differ from those of SP cells. (A) Typical voltage responses (top) to current injections (bottom) recorded in current-clamp mode in an mCit⁺ putative SL cell (left), an mCit⁻ putative SL cell (middle), and an SP cell in deep layer 2 (right). (B) Mean resting membrane potential (E_r) measured in mCit⁺ and mCit⁻ SL cells in 48L mice (bars labeled a and b, respectively), in basal dendrite-lacking and basal dendrite-expressing SL cells in wild-type mice (bars labeled c and d, respectively), and in SP cells in 48L mice (bar labeled e). Error bars indicate SEM. (C–E) Similar plots for the mean input resistance (R_in) (C), mean membrane time constant (τ_m) (D), and mean burst index (a measure of burst firing of APs) (E). NS, not significantly different (P > 0.05); **P < 0.001; *P < 0.05.

We found only minor differences in intrinsic electrical properties between these 4 SL cell variants (Fig. 3B–E). For example, the following mean values were obtained for 48L mCit⁺ and mCit⁻ cells, respectively: resting membrane potential, −62.9 ± 1.5 mV, −65.2 ± 1.1 mV, P = 0.23; input resistance, 284 ± 27 MΩ, 212 ± 26 MΩ, P = 0.07; membrane time constant, 23.2 ± 2.6 ms, 22.4 ± 2.2 ms, P = 0.8; burst index, 1.20 ± 0.06, 1.45 ± 0.15, P = 0.16 (all statistical tests used ANOVA with Tukey's test for multiple comparisons, n = 10 cells in each sample). Note that the trend toward a lower input resistance in mCit⁻ SL cells (212 cf. 284 MΩ, P = 0.07), probably reflects the longer dendrites of mCit⁻ SL cells (Fig. 2C). All electrical properties were similar to the values previously reported for SL cells in wild-type mice (Suzuki and Bekkers 2006, 2011). Finally, all of these properties differed significantly from those measured in layer 2b SP cells in 48L mice (compare bars labeled e, Fig. 3B–E), recapitulating the differences between SL and SP cells reported previously in wild-type mice (Suzuki and Bekkers 2006, 2011).

To summarize the last 2 sections, in 48L mice, mCit labels the approximately 50% of SL cells that entirely lack basal dendrites. Labeled and unlabeled SL cells do not differ in their intrinsic electrical properties.

ChR2 Can Be Selectively Expressed in mCit⁺ SL Cells

Having demonstrated the specificity of tTA-mCit expression in 48L mice, we next used the viral construct AAV-TRE-ChR2-mCherry to enable targeted expression of ChR2 in mCit⁺ SL cells in these animals (Fig. 4A). Typically, for each animal, 1 or 2 slices of the series of coronal slices of the aPC contained the highest number of ChR2-mCherry⁺ neuronal somata, with the number of positive somata falling off steeply (over ∼300–600 µm) in other slices in the series. On the other hand, ChR2-mCherry⁺ putative boutons could be seen at considerable distance (>1500 µm) from the injection site, consistent with the broad extent of associational fibers (Johnson et al. 2000; Franks et al. 2011; data not shown). Comparing the slice(s) with the highest expression from different animals, it was estimated that 11–80% of mCit⁺ SL cells expressed ChR2-mCherry (mean 37.3 ± 9.6%, n = 9 animals analyzed), with the variability likely due to the accuracy with which virus could be injected into the aPC of newborn 48L mice. ChR2-mCherry was very occasionally seen in cells that appeared to lack mCit expression (e.g., Fig. 4A, single red cell located in layer 2b), but this off-target expression was not examined further.

Control experiments confirmed that mCit⁺/mCherry⁺ SL cells responded vigorously to the application of blue light (Fig. 4B, similar results in n = 8 cells). Thus, ChR2 is predominantly expressed in SL cells and is functional, allowing us to use this system to study the postsynaptic targets of a population of SL cells in the aPC.

ChR2-Assisted Circuit Mapping Shows that SL Cells Preferentially Innervate Pyramidal Cells in the aPC

Whole-cell recordings were made from mCherry-negative SL, SP, and DP cells by targeting neurons in layers 2a, 2b, and 3, respectively, of the aPC of virus-injected 48L mice (Fig. 5A). Because associational fibers in the PC extend over long distances (>1000 µm; Haberly and Price 1978; Franks et al. 2011), we recorded from neurons in slices taken from the entire extent of the aPC (>1500 µm along the rostro-caudal axis). Confirmation of the identity of each neuron was obtained by measuring its intrinsic electrical properties in current-clamp mode (as in Fig. 3). Next, TTX (0.5 µM) and 4-AP (100 µM) were washed into the bath to, respectively, block polysynaptic connections and enhance the light-evoked release of neurotransmitter from single ChR2-expressing boutons (Petreanu et al. 2009). Single pulses of blue light (halfwidth 8–10 ms) were then applied across the whole field (diameter ∼400 µm, encompassing layers 1–3) while holding the recorded cell at −70 mV under voltage clamp (Fig. 5D). Because of the presence of TTX, light-evoked EPSCs should only occur in neurons that receive monosynaptic input from ChR2⁺ boutons from SL cells (Fig. 5A). To further check that the light-activated response was not due to the presence of ChR2 in the recorded cell, DNQX (10 µM) was subsequently added to the bath to block AMPA receptors. In every case tested (n = 8 cells), the light-evoked response was abolished, confirming its synaptic origin. Finally, the patch electrode was withdrawn, the slice was fixed, and the identity of the neuron was confirmed by recovering its dendritic morphology.

The light-evoked EPSCs were analyzed as described in Materials and Methods and illustrated in Figure 5B,C. Briefly, the mean charge carried by each light-evoked EPSC was compared with a detection threshold of −70 pC, which corresponds to the upper one-percentile (P = 0.01) of the probability distribution for the charge of the spontaneous background activity measured across all neurons in the dataset (n = 129; Fig. 5C, inset; superimposed gray line is the fit of the extreme value [Gumbel] distribution; vertical dashed line indicates the −70 pC threshold; see Materials and Methods for discussion of other threshold criteria). If the light-evoked EPSC charge exceeded the −70 pC threshold (Fig. 5C, main panel), the cell was scored as receiving monosynaptic input from a ChR2-expressing SL cell. The amplitude of this response was then measured from the peak of the averaged light-evoked current.

For ChR2⁻ SL cells in TTX + 4-AP, none of the 34 cells tested showed a response (Fig. 5D, top panels, examples from 3 SL cells; summary in Fig. 5E; same result for other detection thresholds discussed in Materials and Methods). To test if light-evoked EPSCs would emerge with stronger stimuli, in some cells the flash halfwidth was increased to 50 or 100 ms from the standard 8–10 ms. The longer flash was still unable to elicit a light-evoked EPSC (n = 11 SL cells tested out of the dataset of 34 cells). We also performed a positive control experiment to test our ability to detect light-evoked EPSCs in SL cells under conditions when they are expected to occur. AAV-CMV-ChR2-mCherry was injected into the OB to express ChR2 in mitral/tufted cells, the axons of which provide afferent input to the PC. Three weeks later, slices of the aPC were prepared as usual and light flashes were applied while recording from SL cells. Light-evoked EPSCs were recorded in 4 out of 5 SL cells tested (mean amplitude −70.6 ± 16.5 pA, n = 4, data not illustrated), confirming that we are capable of detecting monosynaptic light-activated EPSCs in these cells, when they are present.

Next, we returned to the original protocol (i.e., AAV-TRE-ChR2-mCherry injected into the aPC) and looked for light-evoked EPSCs in SP and DP cells. In contrast to the results for SL cells, 15 out of 86 SP cells (17.4%) exhibited light-evoked EPSCs in TTX + 4-AP, ranging in amplitude from −4.4 to −130.2 pA (mean −29.5 ± 8.6 pA, n = 15; Fig. 5D, middle panels, and E). In addition, 3 out of 9 DP cells (33.3%) showed light-activated responses in TTX + 4-AP (mean −9.0 ± 2.1 pA, n = 3; Fig. 5D, bottom panels, and E). The use of other detection criteria (Materials and Methods) altered these values somewhat. For example, using a threshold of P = 0.001 (−105 pC) rather than P = 0.01 (−70 pC) reduced the percentage of responsive cells to 14.0% (SP) and 22.2% (DP), but had no effect on the result for SL cells (still 0%).

Pairwise comparisons using a χ² 2 × 2 contingency test confirmed that the number of SL responses was significantly different from those for SP and DP cells (SL–SP, P = 0.016; SL–DP, P < 0.001), but the numbers of SP and DP cells showing responses were not significantly different from each other (SP–DP, P = 0.16; Fig. 5E; analysis for the −70-pC threshold data). Thus, these results indicate that mCit⁺ SL cells monosynaptically excite SP and DP cells but not each other.

A broadly comparable result was obtained for light-evoked EPSCs in the absence of TTX and 4-AP (SL: 0 cells with light-evoked EPSCs out of 26 tested [0%]; SP: 9 out of 32 [28.1%]; DP: 14 out of 25 [56.0%]; all using the −70-pC detection threshold).

Layer-Specific Innervation of SP Cells by SL Cells

In the previous section, we used wide-field light stimulation across all layers of the aPC to estimate the total synaptic input received by ChR2⁻ SL, SP, and DP cells from ChR2⁺ SL cells. Next, we studied the laminar specificity of SL inputs onto SP cells by using a digital projector to restrict the light stimulation to single layers of the aPC (Fig. 6A). Because of the lower light intensity and slow refresh rate of the projector (Materials and Methods), the light-evoked responses obtained in these experiments were weaker and more asynchronous than those obtained with the Hg lamp (Fig. 5), but they did allow comparisons between laminae for the same cell (Fig. 6B).

When recording from ChR2⁻ SP cells, light-evoked EPSCs were much larger when the light stimulus was applied to an associational layer (1b, 2, or 3) than to the afferent layer (1a) (Fig. 6B). Broadly similar results were found in the absence and presence of TTX + 4-AP (black and gray traces, respectively, Fig. 6B). The results were summarized by plotting the mean area under the light-evoked EPSC versus the layer to which light was applied, averaging across all SP cells that showed a response to light stimulation in at least one layer (n = 9; Fig. 6C). The response to light stimulation of layer 1a was significantly smaller than those to stimulation of the other layers (except L3 in the presence of TTX; Fig. 6C, t-test, n = 9 cells, P < 0.03), but there was no significant difference when comparing the responses in layers 1b, 2, and 3 with each other (ANOVA, P > 0.1).

Different Subtypes of GABAergic Interneurons Respond Differently to Light Activation of SL Cells

Finally, we asked whether GABAergic interneurons in the aPC also receive excitatory input from mCit⁺ SL cells and, if so, whether there are any differences between the different classes of interneurons. To help us locate interneurons in live tissue slices, we crossed 48L mice with GAD67-GFP (Δneo) mice in which GABAergic interneurons are labeled with GFP. Double-positive progeny expressed mCit in SL cells in layer 2a and GFP in interneurons in all layers. Although these 2 sources of fluorescence could not be distinguished spectrally, ambiguity in the visual identification of cell type was largely confined to layer 2a where mCit⁺ cells are concentrated; for all other layers, use of these animals was highly advantageous for targeting interneurons. ChR2 was expressed in SL cells of these double-positive animals using viral injection, as before, and whole-cell recordings were made from GFP⁺/ChR2⁻ GABAergic interneurons that were identified by their laminar location, GFP intensity, and intrinsic electrical properties (Fig. 7A), as previously described (Suzuki and Bekkers 2010a, 2012). Control experiments with these animals confirmed that mCit⁺ cells still exhibited the electrical and morphological features of SL cells (data not shown).

Figure 7.

Three main classes of GABAergic interneurons in the aPC receive excitatory input following excitation of ChR2-expressing SL cells when recording in the absence of TTX and 4-AP. (A and B) Each row shows data from a typical example of each of the 3 classes of interneurons that received light-evoked input (FS, RS, and L3 NG cells). Data in each row are from the same cell. (A) Response of each cell to a hyperpolarizing current step (−80 pA) and a depolarizing step close to rheobase (FS, 200 pA; RS, 40 pA; and L3 NG, 160 pA), recorded in current clamp. (B) Typical averaged EPSCs recorded in each cell in response to a wide-field flash of blue light in the absence of TTX and 4-AP. Each trace is an average of 40 episodes. Time scale bar at bottom applies to all traces.

A critical part of the identification of interneurons was the measurement of their firing properties, recorded in the absence of TTX (Fig. 7A). Ideally, monosynaptic light-evoked EPSCs would then be measured in interneurons after perfusing with TTX + 4-AP, as in the recordings from principal cells described earlier. Unfortunately, however, our supply of double-positive (48L/GAD67-GFP) animals was limited, requiring us to record from multiple interneurons in each slice. Because TTX and 4-AP are not easily washed out of slices, all of the experiments described in this section were necessarily performed in the absence of these blockers.

Wide-field light stimulation elicited EPSCs in only 3 classes of interneurons, all of them located in L3: Fast-spiking (FS) cells (9 cells with light-evoked EPSCs out of 13 tested [69.2%], mean EPSC amplitude −70.2 ± 19.8 pA), regular-spiking (RS) cells (2 out of 6 cells [33.3%], −5.8 ± 2.2 pA), and L3 neurogliaform (NG) cells (12 out of 15 cells [80.0%], −46.3 ± 20.1 pA; Fig. 7). No response was observed in layer 1a NG cells (n = 6), layer 1a horizontal (HZ) cells (n = 4), or layer 2 bitufted (BT) cells (n = 9) (data not shown). The percentages of responsive FS and L3 NG cells were both significantly different from those for HZ, BT, and layer 1a NG cells (P < 0.015, χ² 2 × 2 contingency test) while the difference was not significant for RS cells (P > 0.06).

The decay time constant (τ_d) of light-evoked EPSCs was significantly faster in FS cells than in L3 NG cells (FS: τ_d = 3.4 ± 0.6 ms, n = 9; L3 NG: τ_d = 10.9 ± 1.4 ms, n = 8; P < 0.001; Fig. 7B). This difference is reminiscent of that found when using extracellular electrical stimulation to excite associational inputs to these same neurons (FS: τ_d = 3.4 ± 0.3 ms; L3 NG: τ_d = 6.0 ± 0.8 ms; Suzuki and Bekkers 2010a), consistent with the view that light-evoked and electrically evoked EPSCs both arise from associational inputs.

Discussion

In this study, we describe a novel transgenic mouse model (48L) that allows us to use an optogenetics approach to probe the intracortical connectivity of the aPC. Previous work using extracellular electrical stimulation has found that SL cells receive stronger afferent inputs from the OB but much weaker associational (intracortical) inputs than do SP cells (Davison and Ehlers 2011; Franks et al. 2011; Suzuki and Bekkers 2011; Hagiwara et al. 2012). The results presented here give a partial explanation for those earlier observations: We find that mCit⁺ SL cells avoid forming synaptic connections with other SL cells and instead target SP cells, as well as principal cells in L3. This finding suggests that SL cells, in general, may be deficient in a major source of associational inputs—those that originate from other SL cells—and instead contribute most of their output to the powerful intracortical inputs received by SP cells. In addition, mCit⁺ SL cells are selective in their targeting of GABAergic interneurons, providing most of their excitation to FS and NG cells located in L3. These results suggest that SL cells are specialized for providing feedforward excitation of other classes of neurons located in deeper layers of the aPC.

Note that our findings are limited to young mice aged approximately P22–30. Older animals were not tested because of the loss of mCit expression that begins after about P30 in the 48L mouse. It is possible that reciprocal connections between SL cells emerge with age, perhaps because of an age-dependent plasticity mechanism at associational synapses. However, it has been suggested that plasticity at associational synapses in the PC is maintained throughout life, remaining unchanged both during and beyond the early critical period (Franks and Isaacson 2005). Thus, it seems less likely that associational SL–SL cell connections that are completely absent by the end of the critical period (∼P20–30) will emerge later.

The 48L Mouse

Our findings depend on the availability of a new transgenic mouse line, the 48L mouse, in which tTA-mCit is expressed in a subpopulation of SL cells in the aPC. We find that mCit⁺ SL cells comprise about 50% of the total population of SL cells in the upper one-third of layer 2, and that mCit expression is associated with SL cells that entirely lack basal dendrites (Fig. 2). Importantly, about 50% of SL cells in wild-type mice also lack basal dendrites, suggesting that mCit in the 48L mouse is simply reporting pre-existing dendritic diversity and is not a cause of such diversity.

The 48L mouse was developed using an enhancer trap strategy in which a lentiviral vector carrying a single-copy enhancer detector transgene is randomly inserted into the genome of single-cell mouse embryos (Kelsch et al. 2012). Expression of the reporter gene in the detector probe does not faithfully reproduce the expression pattern of the endogenous gene where it integrates, because transcription depends on the interaction between the minimal promoter in the probe and the enhancers neighboring the insertion site. Nevertheless, a simple interpretation of our observations in the 48L mouse is that the transgene is inserted near a region related to some aspect of dendritic extension in SL cells (Sugino et al. 2006; Molyneaux et al. 2007; Tantirigama et al. 2014).

The 48L mouse provides further evidence that principal cells in layer 2 of the PC are not homogeneous. As well as reinforcing the distinction between SL and SP cells (Suzuki and Bekkers 2006, 2011; Hagiwara et al. 2012; McGinley and Westbrook 2013; Fournier et al. 2014), our data also reveal some morphological diversity among SL cells. On the other hand, these different SL cells have similar intrinsic electrical properties (Fig. 3), suggesting that any functional differences will be minor (see next section).

Labeling by mCit is not just restricted to the PC; mCit⁺ cells are found in a number of regions in the brains of 48L mice (Fig. 1). Although glutamatergic neurons in the OB (mitral cells, Fig. 1C) and the hippocampus (granule and pyramidal cells, Fig. 1D) are not labeled, some cells with a pyramidal morphology are labeled in the somatosensory cortex (Fig. 1E). (Interestingly, these neocortical cells do not lack basal dendrites, unlike mCit⁺ SL cells.) Despite this widespread expression of mCit, the labeling in the aPC was always consistent, with the vast majority of labeled cells located in superficial layer 2. However, mCit labeling in the aPC declined with age, consistent with the transgene insertion site being associated with development (Molyneaux et al. 2007).

Connections with Excitatory Neurons

Our main finding was that mCit⁺ SL cells preferentially make intracortical (associational) synaptic connections with SP and DP cells, not with mCit⁻ SL cells. We did not test inputs onto mCit⁺/ChR2⁻ SL cells because, under our recording conditions, these were difficult to identify unequivocally. However, given that mCit⁺ SL cells have even fewer dendrites than mCit⁻ SL cells (Fig. 2), it seems unlikely that the former will receive associational inputs when the latter receive none.

Our finding is in accord with previous reports, using extracellular electrical stimulation, that SL cells receive strong afferent inputs from the OB but weak associational inputs from other neurons within the PC (Suzuki and Bekkers 2006, 2011; Wiegand et al. 2011). This finding is also consistent with the presence of axon collaterals from SL cells in layer 1b (Fig. 2A), which show up as a dense band of labeling in layer 1b in 48L mice (Fig. 1A). Layer 1b is an associational layer of the PC, but the dendrites of SL cells possess few spines within this layer, contrasting with the high spine density on SP cells (Haberly 1983; Neville and Haberly 2004). Assuming that spines are required to receive excitatory synapses, it appears that SL cells are less equipped to receive associational input, whether from other SL cells (as shown here) or from SP or DP cells.

We also mapped the spatial distribution of SL cell inputs onto SP and DP cells by using layer-specific illumination (Fig. 6). The weakest light-evoked inputs were found in layer 1a, consistent with the predominance of afferent (LOT) inputs in this layer (Neville and Haberly 2004), but no significant differences between associational sublaminae were found (Fig. 6C). This finding confirms that associational axons from SL cells are not only concentrated in the dense band in layer 1b noted in the previous paragraph, but also project more broadly throughout deeper layers (Fig. 2A).

We found that the inferred connectivity was similar in both the presence and absence of TTX + 4-AP, suggesting that at least part of the SL–SP and SL–DP connectivity is monosynaptic (Petreanu et al. 2007, 2009). However, the quantification of this connectivity (Fig. 5E) needs to be treated with caution, especially for DP cells, which are undersampled in our dataset. The persistence of light-evoked EPSCs in TTX + 4-AP requires the expression of sufficient ChR2 in presynaptic boutons to locally trigger the release of neurotransmitter, and this might not always be the case (Quattrocolo and Maccaferri 2014). Thus, the percentage of cells showing a light-evoked response in TTX + 4-AP may be an underestimate of the true connectivity if not all ChR⁺ boutons are capable of releasing glutamate.

Connections with Inhibitory Neurons

We found that only 3 classes of interneurons—FS, RS, and NG cells, all of them in L3—received light-evoked EPSCs, with the RS cell response being by far the weakest (although cell numbers for RS cells were low). For technical reasons, these experiments were performed only in the absence of TTX + 4-AP, so it is possible that some of these inputs were polysynaptic. However, dual whole-cell recordings have directly shown that SL cells do make monosynaptic connections with FS cells (Suzuki and Bekkers 2012), raising confidence that our measurement of light-evoked EPSCs in the absence of TTX + 4-AP does correctly report the existence and properties of monosynaptic inputs from SL cells.

Functional Implications

Our findings suggest that SL cells specialize in providing feedforward excitation of SP cells and deep layer interneurons. Intriguingly, this architecture is reminiscent of that in another “old” cortex, the archicortex (hippocampus), in which granule cells in the dentate gyrus generate feedforward excitation of CA3 pyramidal cells. Dentate granule cells even have a striking physical resemblance to SL cells, both of them being deficient in basal dendrites. It has been suggested that dentate granule cells might remove redundancy from inputs to the hippocampus before information is passed on to the autoassociative network of CA3 pyramidal cells, where memories are represented (Treves and Rolls 1994). SP cells in the PC have long been modeled as an autoassociative network for the sparse representation of odor memories (Haberly and Bower 1989; Hasselmo and Barkai 1995; Hopfield 1995). It is interesting to speculate about whether SL cells and dentate granule cells have evolved similar preprocessing roles in their respective cortices.

In conclusion, we report here the first use of a new transgenic mouse, the 48L mouse, to examine the postsynaptic targets of SL cells in the aPC. Our results provide new information about the connectivity of the aPC, and provide further evidence for the functional specialization of the different classes of principal cells in the main input layer of this sensory paleocortex.

Funding

This work was supported by the National Health and Medical Research Council of Australia (471413 and 585462 to J.M.B.); the National Eye Institute (EY022360 to S.B.N.); the National Institute of Neurological Disorders and Stroke (to S.B.N.); and the Graduate University of Advanced Studies, Okazaki (Short-Stay Abroad Program Grant to T.B.).