The olfactory system of teleost fish sends the unbranched OSN axon to a single glomerulus (Weiss et al., 2020). The layer structure of the teleost fish OB is similar to that of the rodent OB, but a bit more simplistic (Kermen et al., 2013). In the fish OB, there are two types of glutamatergic projection neurons; mitral cells, a major projection neuron, and ruffed cells (Figure 2A1). There is no clear MCL, and both mitral cells and ruffed cells are located in the external cell layer that lies beneath the GL (Satou, 1990). Although they share the same name and have apical dendrites innervating glomeruli allowing them to receive input from OSN axons, the mitral cells of teleost fish are significantly different from those in the mammalian OB. Golgi staining and retrograde tracing of OB neurons in adult teleosts revealed that mitral cells do not extend secondary dendrites (Kosaka and Hama, 1982; Fuller et al., 2006). In most teleost fish, mitral cells possess multiple apical dendrites that project to multiple glomeruli, with the exception of zebrafish (Kosaka and Hama, 1982; Oka, 1983; Fujita et al., 1988; Satou, 1990; Fuller et al., 2006). The zebrafish mitral cells typically have only a single apical dendrite innervating a single glomerulus, and a smaller percentage of mitral cells have multiple apical dendrites that still innervate a single glomerulus (Fuller et al., 2006). In contrast to the mitral cells, the ruffed cells have a membranous field surrounding the initial part of the axon (Kosaka and Hama, 1979; Fuller and Byrd, 2005). A major difference from mitral cells is that ruffed cells do not receive direct input from OSN axons, but receive inhibitory synaptic inputs from interneurons activated by mitral cells (Kosaka, 1980; Satou, 1990). Therefore, the activity patterns of mitral and ruffed cells are contrasting in nature (Zippel, 1999; Zippel et al., 2000).

Mitral and ruffed cells project their axons to higher brain regions through either the medial olfactory tract (MOT) or LOT depending on the location of their somata (Fuller et al., 2006; Miyasaka et al., 2014). In the brain, mitral cell axons send their projections to five major regions; posterior and ventral telencephalon, posterior tuberculum, right habenula, and ipsilateral/contralateral OB (Eisthen and Polese, 2010) (Figure 2A2). Genetic labeling of zebrafish mitral cells at single-cell resolution revealed that (1) individual mitral cells can target multiple regions; the MOT and LOT do not determine the target regions, (2) mitral cells innervating the same glomerulus do not show the same axon trajectory and (3) the right habenula is innervated by mitral cells that convey the information from the medial glomeruli (Miyasaka et al., 2009, 2014).

Although there is no distinct AOB in the teleost OB, a report suggested that a specific glomerulus located at the mediodorsal OB, mdG2, may be functionally similar and serve as the accessory olfactory system in teleost fish (Biechl et al., 2017). The mdG2 receives inputs solely from crypt cells, a specific type of OSN activated by kin odor, and the mitral cells targeting the mdG2 send their axons to the intermediate ventral telencephalic nucleus (Ahuja et al., 2013; Biechl et al., 2016, 2017; Gerlach et al., 2019). It is hypothesized that this nucleus in teleost fish may be functionally analogous to the medial amygdala of mammals (Biechl et al., 2017). The evolutionary origin of the accessory olfactory system is an intriguing topic to aid in our understanding of the parallel pathways for olfactory information processing. The axonal projection patterns of the ruffed cells in the fish olfactory system is another research topic that remains to be elucidated.

In contrast to the mammalian olfactory system, a single OSN of amphibians typically innervates multiple glomeruli (Gilbert et al., 2013; Hassenklover and Manzini, 2013). The OB of the frog has a layer structure similar to, but not as fully concentric as, that of the mammalian OB (Byrd and Burd, 1991; Scalia et al., 1991a; Kratskin et al., 2000). The projection neurons called mitral cells are scattered among the EPL and MCL (Kratskin et al., 2000). The basic morphology of the frog mitral cells resembles that of rodent mitral and tufted cells, except for their multiple (1–6) primary dendrites that innervate multiple glomeruli (Jiang and Holley, 1992) (Figure 2B1). Jiang and Holley (1992) further demonstrated that the mitral cells located superficially, close to the GL, extend their primary dendrites with a larger angle and targeted glomeruli across a wider field of range. The primary and secondary dendrites of the frog mitral cells are not clearly distinct in that some dendrites send several branches into the GL while the others remain in the EPL (Jiang and Holley, 1992). Mitral cells with similar morphology were also observed in the salamander OB (Herrick, 1924). A unique feature of amphibians is the process of metamorphosis in which species transform from an immature to mature state. This is an intriguing process with respect to olfactory system development and evolution. The studies of the OB morphology in the clawed frog showed that the basic structure of the mature OB was apparent in tadpoles (around larval stage 48/49) and remained constant throughout the late larval stage and into adulthood, with only the size increasing (Byrd and Burd, 1991; Nezlin and Schild, 2000). The majority of MOB projection neurons in both frog and salamander tadpoles (stages 51–56) had more than one glomerular tuft (up to 4) innervating different glomeruli (Herrick, 1924; Nezlin et al., 2003).

As shown in Figure 2B2, OB projection neurons project to the dorsal, lateral, and medial pallium as well as the lateral and medial septal nuclei in the frog brain (Eisthen and Polese, 2010). They also innervate the lateral amygdala (Scalia et al., 1991b). Contrary to the lack of an apparent AOB in the teleost fish, the AOB in the frog is quite evident and is located at the ventrolateral region of the OB. Anterograde HRP tracing experiments revealed that the AOB projection neurons project to the medial amygdala (Scalia et al., 1991b). While the projection from the MOB and AOB may converge in the amygdala, the medial amygdaloid nucleus may be connected exclusively to the AOB (Scalia et al., 1991b) (Figure 2B2).

Whether a single OSN of reptiles projects unbranched axon to a single glomerulus has not been clearly described. The MOB of reptiles also exhibits a distinct layer structure in which the GL, EPL, MCL, and GCL are clearly visible, however, they may have a thicker IPL than mammalian OB (Kirillova and Lin, 1998; Pinato and Midtgaard, 2003; Kondoh et al., 2013). Mitral cells are distributed throughout the MCL in the turtle OB (Kirillova and Lin, 1998). The morphology of turtle mitral cells was determined by reconstruction after electrophysiological recordings and showed the existence of long secondary dendrites that reach almost half of the bulbar circumference (Mori et al., 1981). In the turtle and snake OB, a single mitral cell possesses more than one primary dendrite and therefore is capable of targeting multiple glomeruli (Mori et al., 1981; Iwahori et al., 1989) (Figure 2C1). However, mitral cells in the lizard OB have only one dendritic tuft in a glomerulus, which was shown with Golgi staining (Llahi et al., 1985) (Figure 2C1). While not yet experimentally concluded, the existence of tufted cells in the reptile OB is suggested due to the discernible segregation of the EPL and MCL. In fact, the mitral cells observed in the EPL of the lizard OB are described as displaced mitral cells (Llahi et al., 1985). Cells varying in soma size were identified in the EPL and MCL of the snake MOB (Kondoh et al., 2013), and the Japanese striped snake has two mitral cell types that are morphologically distinct based on the somata locations within the MCL (Iwahori et al., 1989). These support a concept suggestive of a transition between mitral cells and tufted cells.

In the snake brain, the axonal projections from the MOB terminate at the AON, olfactory tubercle (OT), the lateral cortex, and some amygdaloid nuclei (Halpern, 1976; Martinez-Garcia et al., 1991; Lohman and Smeets, 1993; Lanuza and Halpern, 1998; Eisthen and Polese, 2010; Ubeda-Banon et al., 2011) (Figure 2C2). Overall, the projection patterns of most reptilian MOB projection neurons are somewhat comparable to that of the rodent (see Figures 2C2, 4). However, in contrast to the mammalian OB projection neurons, some axons from the snake MOB enter the ipsilateral stria medullaris thalami, cross the midline in the habenular commissure, enter the contralateral stria medullaris thalami and terminate in the contralateral lateral pallium (Halpern, 1976; Lanuza and Halpern, 1998). The accessory olfactory system becomes more noticeable in reptiles compared to amphibians considering that the AOB is caudally located and clearly separated from the MOB (Martínez-Marcos and Halpern, 2009; Ubeda-Banon et al., 2011; Kondoh et al., 2013). The AOB mitral cells in the reptile give rise to more than one primary dendrite with multiple tufts in the GL (Llahi et al., 1985). The axons of AOB mitral cells follow along the accessory olfactory tract and project to three portions of the amygdala: the nucleus sphericus, medial amygdala, and nucleus of the accessory olfactory tract (Halpern, 1976; Martinez-Garcia et al., 1991; Lohman and Smeets, 1993; Lanuza and Halpern, 1998; Eisthen and Polese, 2010; Ubeda-Banon et al., 2011) (Figure 2C2). In some reptilian species including a type of lizard, the AOB projects to the bed nucleus of stria terminalis (BNST) (Martinez-Garcia et al., 1991).

Different species have morphologically distinct MOB projection neurons. As described above, the mitral cells of teleost fish, amphibians, and reptiles, excluding zebrafish and lizards, typically possess multiple primary dendrites projecting to multiple glomeruli while only a single primary dendrite is formed in the mammalian mitral cells (Kosaka and Hama, 1982; Oka, 1983; Fujita et al., 1988; Satou, 1990; Dryer and Graziadei, 1994; Fuller et al., 2006). These results suggest that OB projection neurons have reduced the number of their primary dendrites from multiple to a single primary dendrite over the course of evolution. Interestingly, rodent MOB mitral cells form multiple primary dendrites with tufts in multiple glomeruli at the initial stage of development (Malun and Brunjes, 1996; Lin et al., 2000). Although it is reasonable to assume that having a single primary dendrite projecting to a single glomerulus assists in odor discrimination, it is still largely unknown how each mitral cell is able to “select” one dendrite during maturation. A recent study showed that spontaneous network activity among immature projection neurons in the neonatal OB is essential for the pruning of excess primary dendrites, but the OSN activity does not appear to be necessary in this process (Lin et al., 2000; Fujimoto et al., 2019). Comparing the development of OB projection neurons in different species would provide us with interesting insights into the molecular and cellular mechanisms underlying the selection of a single primary dendrite.

It is also noteworthy that clear secondary dendrites seem to be first apparent in the reptilian mitral cells, whereas the distinction between the primary and secondary dendrites in amphibian mitral cells is difficult to observe. From an evolutionary perspective, this information suggests that information processing in the OB became more intricate and complex as life began to shift from living in water to living on land. As discussed in the latter section, difference in the length of secondary dendrites is important for OB projection neurons to differentially respond to the odor inputs in the rodent OB. Formation of the two characteristic dendrites might be a stepping stone to generate the parallel processing pathways in the olfactory system.

In the rodent OB, the projection neurons located in the MCL are defined as mitral cells, and others found in the EPL and GL are deemed tufted cells (Ramon y Cajal, 1911; Shepherd et al., 2004; Greer et al., 2008; Ennis et al., 2015). However, based on the relative depth of somata location in the EPL and GL, tufted cells are further subclassified into the external, middle, and internal tufted cells, also known as superficial, intermediate, and deep tufted cells, respectively (Ramon y Cajal, 1911; Macrides and Schneider, 1982; Orona et al., 1984; Shepherd et al., 2004). The internal tufted cells are sometimes identified as displaced mitral cells because of their proximity to the MCL (Mori et al., 1983; Kishi et al., 1984; Shepherd et al., 2004). The size of mitral cell somata (20–25 μm) is typically larger than that of tufted cells (10–20 μm) (Pinching and Powell, 1971a). Both mature mitral and tufted cells have a single primary dendrite with a tuft at the tip residing in a single glomerulus where they receive excitatory input from OSN axons (Ramon y Cajal, 1911; Mori et al., 1983). Mitral and tufted cells also form dendrodendritic synapses with inhibitory periglomerular interneurons in the GL (Pinching and Powell, 1971b).

In the rodent OB, different types of projection neurons possess different lengths of secondary dendrites. Mori et al. studied projection neurons in the rabbit OB, and showed that a single mitral cell possesses ∼15,000 μm of secondary dendrites, which is almost four times longer than that of middle tufted cells (∼4,000 μm) (Mori et al., 1983). Internal tufted cells have an intermediate length of secondary dendrites (∼12,000 μm) (Mori et al., 1983). Also in the mouse OB, it was shown that the total dendritic length of a mitral cell (∼15,000 μm) is much longer than that of a middle tufted cell (∼7,500 μm) (Igarashi et al., 2012). The tufted cells located in the most superficial EPL or GL are classified as the external tufted cells. Like the other mitral and tufted cells, the primary dendrites of external tufted cells are generally uni-glomerular, with a small subpopulation being di-glomerular (Ennis and Hayar, 2008). There are two distinct subpopulations of external tufted cells, and the most prominent dissimilarity between the populations is the presence or absence of secondary dendrites (Macrides and Schneider, 1982; Hayar et al., 2004a, b; Antal et al., 2006; Liu and Shipley, 2008; Hirata et al., 2019). The somata of external tufted cells with secondary dendrites are generally found in the deeper one-third of the GL, or in the EPL near the boundary with the GL. Morphometric analysis using 300 μm thick rat OB slices showed that the total length of secondary dendrites of external tufted cells in the slice was ∼1,200 μm (Antal et al., 2006). Although we cannot directly compare the results due to the differences in species and method, it is reasonable to assume that mitral cells possess the longest secondary dendrites while those of the external tufted cells are the shortest. Interestingly, the other group of external tufted cells whose somata are only found in the GL lack secondary dendrites (Macrides and Schneider, 1982; Hayar et al., 2004a, b; Antal et al., 2006; Liu and Shipley, 2008; Hirata et al., 2019). Based on the data acquired thus far, it can be asserted that the deeper the location of projection neurons cell somata in the OB, the longer the secondary dendrites are as shown in Figure 3A.

In addition to the differences in length, Figure 3A shows that mitral and tufted cells extend their secondary dendrites in different sublayers of the EPL. The secondary dendrites of mitral cells appear to remain restricted to the deepest portion of the EPL (dEPL), proximal to the MCL, while those of tufted cells extend in the most superficial portion of the EPL (sEPL), proximal to the GL (Macrides and Schneider, 1982; Mori et al., 1983; Orona et al., 1984; Imamura and Greer, 2015). Even among tufted cells, it has been shown that the external tufted cells extend secondary dendrites to the most outer region of the sEPL and the middle tufted cells to the inner region of the sEPL (Macrides and Schneider, 1982; Mori et al., 1983). Several studies defined an intermediate EPL (iEPL) between sEPL and dEPL, where the secondary dendrites of internal tufted cells are projected (Orona et al., 1984; Mouradian and Scott, 1988). As discussed in a later section, there are mitral cells that extend secondary dendrites to the iEPL or the inner sEPL (Orona et al., 1984). Therefore, the regions in the EPL occupied by secondary dendrites of mitral cells and tufted cells gradually shift from deep to superficial. Interestingly, granule cells are also subgrouped into at least three populations based on their dendritic extension patterns in the EPL: the type-I granule cell ramifies spiny dendrites at any depth of the EPL; dendrites of the type-II granule cell extend only in the deep EPL; and the type-III granule cell ramifies spiny dendrites predominantly in the superficial EPL (Schneider and Macrides, 1978; Mori et al., 1983; Orona et al., 1983; Merkle et al., 2007). A computational analysis indicated that lateral inhibition mediated by dendrodendritic synapses between secondary dendrites of projection neurons and granule cell dendrites in the EPL could spread through granule cells only in a mitral-mitral or tufted-tufted way, but not mitral-tufted (Cavarretta et al., 2018). However, it was also shown that individual granule cells could influence a large group of both mitral and tufted cells belonging to at least 15 glomerular modules (Arnson and Strowbridge, 2017).

The activity of OB projection neurons is also regulated by other interneurons, such as short-axon cells (Pressler and Strowbridge, 2006; Parrish-Aungst et al., 2007; Eyre et al., 2008; Eyre et al., 2009; Arenkiel et al., 2011; Burton, 2017; Burton et al., 2017). In the EPL, mitral and tufted cells form reciprocal and non-reciprocal connections with EPL-located interneurons (EPL-IN) expressing parvalbumin (PV), somatostatin (SST), vasoactive intestinal peptide (VIP), and/or Corticotropin-Releasing Hormone (CRH) (Toida et al., 1994; Lepousez et al., 2010; Huang et al., 2013; Matsuno et al., 2017). EPL-INs form broader patterns of connectivity with mitral and tufted cells than granule cells (Huang et al., 2016), which is consistent with the broader odor tuning of EPL-INs (Kato et al., 2013; Miyamichi et al., 2013). Interestingly, CRH-positive EPL-INs provide stronger inhibition onto tufted cells than mitral cells, and tufted cells exhibit more linearly additive responses to odor mixtures without EPL-IN inhibitions (Liu G. et al., 2019). On the other hand, SST-positive EPL-INs are located in the dEPL and extend dendrites specifically into the dEPL (Lepousez et al., 2010). Together with the results showing that lateral inhibition differs between mitral cells and tufted cells (Geramita et al., 2016; Geramita and Urban, 2017; Matsuno et al., 2017), these results suggest that mitral and tufted cell activities are regulated mostly by different inhibitory circuits.

Axons of OB projection neurons ramify both within the OB and in the olfactory cortex (Kishi et al., 1984; Ojima et al., 1984; Orona et al., 1984; Igarashi et al., 2012). Within the OB, the axon collaterals terminate predominantly in the GCL to form asymmetric synapses on somata and dendrites of granule cells and short axon cells (Price and Powell, 1970a, b; Eyre et al., 2008). In the rabbit OB, the collaterals of mitral cells were distributed widely from the deep portion to the most superficial portion of the GCL (Kishi et al., 1984). The collaterals of internal tufted cells tended to be distributed more superficially in the GCL than mitral cells while those of middle tufted cells were distributed in the most superficial GCL (Kishi et al., 1984). It was also reported that the collaterals of external tufted cells run through the IPL to connect lateral and medial sides of the odor map (Liu and Shipley, 1994; Belluscio et al., 2002; Lodovichi et al., 2003). Since granule cells that project their dendrites to the superficial and deep EPL tend to localize in the superficial and deep GCL, respectively (Orona et al., 1983; Imamura et al., 2006), different types of OB projection neurons seem to construct distinct neuronal microcircuits within the OB.

Projection neuron axons extend from the ventrolateral side of the OB and form the LOT before innervating the olfactory cortex (Kishi et al., 1984; Yamatani et al., 2004; Walz et al., 2006; Igarashi et al., 2012). AOB mitral cell axons pass through the deepest layer of the LOT, and the axons of the MOB mitral and tufted cells are found in the intermediate and superficial layers, respectively (Inaki et al., 2004; Yamatani et al., 2004). The location of the axons assists in preserving the topographical organization of the olfactory information as it extends from the OB to the olfactory cortex. Target regions of the AOB mitral cells and MOB mitral and tufted cells rarely overlap (Figure 4). The AOB mitral cells transmit the information from the vomeronasal organ to the bed nucleus of the accessory olfactory tract (BAOT), the BNST, the medial amygdaloid nucleus (MEA), and the posteromedial cortical amygdaloid nucleus (PMCo) (Scalia and Winans, 1975; Davis et al., 1978; Yoshihara et al., 1999; von Campenhausen and Mori, 2000). The MOB mitral and tufted cells innervate the olfactory cortex consisting of the AON, the anterior and posterior piriform cortex (aPC and pPC), the OT, the lateral entorhinal cortex (LEC), the MEA, and the anterior and posterolateral cortical amygdaloid nucleus (ACo and PLCo) (Scalia and Winans, 1975; Haberly and Price, 1977; Yoshihara et al., 1999; Miyamichi et al., 2010; Ghosh et al., 2011; Sosulski et al., 2011; Hintiryan et al., 2012; Igarashi et al., 2012; Hirata et al., 2019). Within the MEA, AOB mitral cell axons terminate in the deep region, and the MOB mitral cell axons are found in the superficial layer without overlap (Kang et al., 2009). Several studies showed that a single mitral cell innervates the entire olfactory cortex while a tufted cell projects only to part of the AON and OT (Nagayama et al., 2010; Igarashi et al., 2012; Hirata et al., 2019). Within the OT, middle tufted cells project to the lateral portion, whereas the medial portion is preferentially innervated by mitral cells (Igarashi et al., 2012). Although it was previously undetermined if external tufted cells project their axons outside the OB (Schoenfeld et al., 1985; Tobin et al., 2010; Lukas et al., 2019), Hirata et al. (2019) recently showed that at least a subpopulation of external tufted cells do target the anterolateral edge of the OT as well as the pars externa of the AON.

Odor responses of mitral and tufted cells are regulated by their intrinsic physiological properties and the intraglomerular and interglomerular circuitry (Aungst et al., 2003; Shao et al., 2012, 2013; Adam et al., 2014; Fukunaga et al., 2014). Many studies have reported the dissimilarity in odor responses between mitral and tufted cells. First, a distinct combination of odorants excites or suppresses the output of mitral and tufted cells. Interestingly, a specific odor can inhibit mitral cell excitation, while the external tufted cells show a distinct lack of the odor-specific suppression which is believed to be shaped by interglomerular inhibition (Adam et al., 2014; Kollo et al., 2014; Banerjee et al., 2015; Economo et al., 2016). Molecular receptive range (MRR) is defined as the odor spectrum that activated by a given odorant receptor, glomerulus, or a mitral or tufted cell, and is important for odor discrimination. It was reported that middle tufted cells have a broader MRR than mitral cells (Nagayama et al., 2004; Adam et al., 2014). Similarly, even among the projection neurons targeting the same glomerulus, the deeper the somata location, the narrower the MRRs (Kikuta et al., 2013). Since the MRRs of OB projection neurons are largely regulated by lateral inhibition mediated by the dendrodendritic synapses formed between mitral/tufted cells and granule cells (Yokoi et al., 1995; Tan et al., 2010; Geramita et al., 2016), the length of the secondary dendrites may be a critical determinant of the MRR.

It has been suggested that not only the MRRs but also the temporal components of the projection neuron activity contributes to odor identification (Schaefer and Margrie, 2007; Uchida et al., 2014; Wilson et al., 2017; Blazing and Franks, 2020). There are notable differences between mitral and tufted cells in temporal activation patterns after odor stimulation. Tufted cells can respond to lower concentrations (∼10 times lower than mitral cells) of odor stimuli with a higher frequency (>100 Hz), whereas the typical firing rate of mitral cells is less than 100 Hz (Nagayama et al., 2004; Igarashi et al., 2012; Kikuta et al., 2013; Adam et al., 2014). In vitro studies suggested that the greater excitability of tufted cells is caused by stronger afferent excitation, greater intrinsic excitability, and less inhibitory tone (Schneider and Scott, 1983; Burton and Urban, 2014; Arnson and Strowbridge, 2017; Geramita and Urban, 2017). On the other hand, mitral cells respond to strong OSN stimulation with sustained firing, or persistent discharge, that continues after odor stimulation (Adachi et al., 2005; Matsumoto et al., 2009; Geramita and Urban, 2017; Vaaga and Westbrook, 2017). The timing of firing onset in reference to the respiratory cycle is also different between mitral and tufted cells. Tufted cell spiking is phase-locked to OSN stimulation without sustained firing and starts during the middle of the inhalation phase (early-onset), while mitral cells respond with later-onset during the transition phase from inhalation to exhalation in anesthetized freely breathing rodents (Fukunaga et al., 2012; Igarashi et al., 2012). However, in an artificial inhalation paradigm, superficial, middle, and deep projection neurons were not reliably distinguished based on the timing of their inhalation-evoked activity (Diaz-Quesada et al., 2018; Short and Wachowiak, 2019).

External tufted cells receive direct OSN input and provide feedforward excitation to other neurons in the GL including periglomerular and short-axon cells and therefore are involved in interglomerular suppression of other OB projection neurons (Aungst et al., 2003; Hayar et al., 2004a; Whitesell et al., 2013; Liu and Liu, 2018). In addition, as described in the previous section, at least a subset of external tufted cells target their axons to the anterolateral edge of the OT and the pars externa of the AON (Hirata et al., 2019), suggesting that they contribute to parallel pathways of the olfactory system. Focusing on intrinsic physiological properties, the external tufted cells inherently generate rhythmic theta bursts (1–10 Hz) of action potentials and respond optimally to rhythmic, sniffing-related input (Hayar et al., 2004b; Liu and Shipley, 2008). On the other hand, mitral cells have biphasic membrane potentials that control the responsivity to OSN stimuli (Heyward et al., 2001; Kollo et al., 2014). As suggested from the differences in intrinsic properties, responses to odor stimuli of external tufted cells are distinct from mitral cells (Vaaga and Westbrook, 2016, 2017). Moreover, cholecystokinin (CCK) is a neuropeptide that is known to express strongly in a subset of the external tufted cells (Seroogy et al., 1985; Liu and Shipley, 1994; Gutierrez-Mecinas et al., 2005; Baltanas et al., 2011), although in situ hybridization analysis and recent immunohistochemical studies indicate a weak CCK expression also in mitral cells (Ingram et al., 1989; Hirata et al., 2019). Optical imaging of different mouse OB projection neurons showed that CCK-positive external tufted cells exhibited a shorter range of odor response latencies and durations than mitral cells and other external tufted cell populations (Short and Wachowiak, 2019). Thus, external tufted cells likely transmit the olfactory information to specific regions in the olfactory cortex with unique temporal patterns. On the other hand, vasopressin, a neuropeptide, is predominantly expressed by external tufted cells with secondary dendrites and some middle tufted cells, but not by mitral cells, in the rodent OB (Tobin et al., 2010; Lukas et al., 2019). It is proposed that the vasopressin-positive external tufted cells are involved in mechanisms of social recognition via the odor signatures (Dluzen et al., 1998a, b; Tobin et al., 2010; Wacker et al., 2011). However, recent studies showed that OSN stimulation primarily caused strong inhibition of the vasopressin-positive external tufted cells, suggesting that OSN axons do not directly activate them (Lukas et al., 2019). Whether the vasopressin-positive external tufted cells form a distinct neural circuit/pathway from the other external tufted cells is needed to be further elucidated.

The various types of oscillations of local field potential (LFP) are produced in the OB after odor stimulation, and they are associated with odor perception, discrimination, and learning (Martin et al., 2004; Beshel et al., 2007; Kay et al., 2009; Lepousez and Lledo, 2013; Martin and Ravel, 2014; Li et al., 2015; Liu P. et al., 2020; Losacco et al., 2020). A previous report showed that the local circuits produce fast-frequency (65–100 Hz) and slow-frequency (35–65 Hz) gamma oscillations of LFP in the OB (Kay, 2003). Several reports suggested that the early-onset fast gamma-oscillations and later-onset slow gamma oscillations are generated mainly by tufted cell and mitral cell subsystems, respectively (Manabe and Mori, 2013; Frederick et al., 2016). On the other hand, it was shown that glomerular networks coordinate theta oscillations (2–12 Hz) (Hayar et al., 2004b; Fukunaga et al., 2014). These findings imply that synchronized oscillatory activity at different frequency may be a key mechanism for OB projection neurons to process the different aspects of odor information in parallel.

Since the MCL is only about the size of one cell body in thickness, heterogeneity of mitral cells has not been deeply investigated. However, several lines of evidence indicate that mitral cells do consist of heterogeneous subpopulations. Although, as noted above, mitral cells usually extend their secondary dendrites in the dEPL (Mori et al., 1983; Orona et al., 1984), some mitral cells extend their secondary dendrites in the iEPL in the rat OB, even though their somata lay in the MCL (Orona et al., 1984; Mouradian and Scott, 1988). Orona et al. (1984) classified the former mitral cells as Type I and the latter as Type II mitral cells (Figure 3A). Interestingly, the total secondary dendrite length of Type II mitral cells was longer than that of the middle tufted cell but shorter than Type I mitral cells (Orona et al., 1984). It remains to be seen whether there is a difference in axonal projection patterns between Type I and Type II mitral cells.

Mitral cells may also be subclassified based on their location along different planes of the OB. For example, mitral cells located at different regions along the dorsomedial-ventrolateral axis in the MCL tend to exhibit different projection patterns toward the OT, cortical amygdala, and MEA (Haberly and Price, 1977; Scott et al., 1980; Miyamichi et al., 2010; Imamura et al., 2011; Inokuchi et al., 2017). Retrograde labeling of OB projection neurons from the olfactory cortex revealed that the cortical amygdala (ACo and PLCo) and OT receive afferent projections preferentially from mitral cells in the dorsomedial and ventrolateral MCL, respectively (Haberly and Price, 1977; Scott et al., 1980; Miyamichi et al., 2010; Imamura et al., 2011). MOB mitral cells that project to the MEA are locally found at the ventral region of the OB and mediate odor-induced attractive social responses (Lin et al., 2007; Thompson et al., 2012; Inokuchi et al., 2017; Lemons et al., 2017).

In contrast to the studies reporting differences between mitral and tufted cells, only a few studies have suggested heterogeneous physiological and molecular properties among mitral cells. It has been shown that the α3 subunit of the GABA_A receptor, as well as a subunit of voltage-gated potassium channel (Kv1.2), are expressed by subsets of mitral cells (Panzanelli et al., 2005; Padmanabhan and Urban, 2010). In addition, a subunit of hyperpolarization-activated cyclic nucleotide-gated channel, HCN2, was expressed in glomeruli in a mosaic pattern (Angelo and Margrie, 2011; Angelo et al., 2012). These studies also reported on the diversity of intrinsic biophysical properties among mitral cells, such as firing frequency and the I_h sag current, which are supposedly reflective of varying Kv1.2 and HCN2 expression levels (Padmanabhan and Urban, 2010; Angelo et al., 2012). Moreover, these differences in molecular and biophysical properties may endow mitral cells with different odor response properties (Dhawale et al., 2010; Kikuta et al., 2013).