Collectively defined by their synthesis of the neurotransmitter norepinephrine, central noradrenergic neurons are heterogeneous. In addition to varying in their anatomical location, morphology, axonal projection pattern, electrophysiological characteristics, and gene expression (Dahlström and Fuxe, 1964; Swanson, 1976; Holets et al., 1988; Robertson et al., 2013; Chandler et al., 2014; Li et al., 2016), noradrenergic neurons exhibit differential susceptibility to neurodegenerative disease and toxicant exposure (German et al., 1992; Mohideen et al., 2011; Theofilas et al., 2017). Determining the molecular basis of this heterogeneity is therefore likely to have important clinical implications, in addition to shedding light on how noradrenergic neurons regulate behaviors and physiological functions as diverse as attention, emotion, appetite, memory, and response to stress (Berridge and Waterhouse, 2003).

The organization of central noradrenergic neurons was initially defined over fifty years ago by their rostrocaudal position in the adult brainstem (Dahlström and Fuxe, 1964). This mature anatomical organization is still the primary means by which the noradrenergic system is studied, but evidence of heterogeneity within anatomically defined noradrenergic nuclei indicates that other approaches will be required to fully explain functional diversity. An alternative strategy for understanding this heterogeneity is the study of genetic neuroanatomy, correlating gene expression with brain structure and function (Joyner and Sudarov, 2012). To this end, we have recently used a recombinase-mediated intersectional genetic fate mapping approach to subdivide the mature noradrenergic system based on gene expression differences along the anteroposterior axis of the embryonic hindbrain (Robertson et al., 2013, 2016). We identified genetically defined noradrenergic subpopulations originating in different rhombomeres (r), transient segments of the embryonic hindbrain with distinct gene expression profiles (Krumlauf et al., 1993; Chambers et al., 2009). The population arising from the isthmus (r0) and r1, as defined by early expression of En1, comprises the majority of the locus coeruleus (LC), the largest anatomically defined noradrenergic population and a major source of norepinephrine released in the cortex, as well as part of the dorsal subcoeruleus (SubC) and A7 noradrenergic nuclei in the pons. More recently, fate mapping on the basis of Fgf8^Cre expression in r0 has demonstrated that the LC is divided between r0- and r1-derived subpopulations (Watson et al., 2017). A dorsal subpopulation is derived from the Fgf8^Cre expression domain encompassing r0, and by inference, the remaining ventral subpopulation is derived from r1.

In addition to the anteroposterior axis, developmental gene expression in the embryonic neural tube also varies on the dorsoventral axis (Echelard et al., 1993; Liem et al., 1995, 1997; Moreno-Bravo et al., 2014; Di Bonito and Studer, 2017). This second axis of developmental gene expression is likely a source of functional heterogeneity in the mature LC and adjacent noradrenergic neuron populations. The developmental progression of Phox2a and Phox2b expression during LC development suggests that the neurons of the LC originate in the alar plate, the dorsal subdivision of the neural tube, and migrate to their final position in the basal plate (Aroca et al., 2006). In chick, the alar origin of the LC was confirmed experimentally by fate mapping with quail/chick chimeras and immunohistochemical analysis using Pax7 as a marker of the alar plate (Aroca et al., 2006). The dorsoventral origins of SubC and A7 noradrenergic neurons derived from the En1 expression domain were not assessed.

In the current study, we use an intersectional genetic fate mapping strategy in mice to directly assess the dorsoventral origins of all the noradrenergic neurons derived from the En1 expression domain encompassing r0 and r1. We confirm that the majority of LC neurons originate in the alar plate, as defined by expression of Pax7^Cre, and demonstrate that this region is also the source of a large proportion of r0- and r1-derived SubC and A7 cells. Surprisingly, we also find that a small subpopulation of noradrenergic neurons distributed throughout the mature LC, SubC, and A7 does not have a history of Pax7^Cre expression, indicating an origin outside the Pax7-defined alar plate. Consistent with the hypothesis that different developmental origins imply different functions, this subpopulation has a distinct axonal projection pattern.

Using intersectional genetic fate-mapping, we have previously shown that noradrenergic neurons derived from the En1 expression domain of the embryonic hindbrain populate 99.8% of the locus coeruleus (LC) and a portion of the dorsal subcoeruleus and A7 nuclei (Robertson et al., 2013) according to an adult mouse brain atlas (Paxinos and Franklin, 2013). These neurons (hereafter designated the LC complex) form a continuum extending from the compact LC proper to the more dispersed cells of the SubC and A7 nuclei (Plummer et al., 2015). An earlier study using in situ hybridization of chick and mouse embryos and quail/chick grafting experiments demonstrated that the LC is derived from the alar plate (Aroca et al., 2006), but these embryonic analyses did not distinguish the LC proper from the SubC and A7, structures more clearly delineated in the adult brain. To determine if all noradrenergic neurons of the LC complex originate from the alar plate, we therefore employed a recombinase-based intersectional and subtractive genetic fate mapping strategy that allows labeled neurons to be mapped into the adult brain (Jensen and Dymecki, 2014).

Mapping noradrenergic neurons derived from the alar plate of r0 and r1 required three recombinase driver alleles and a triple recombinase-responsive indicator allele. Dbh^Flpo (Robertson et al., 2013) is a noradrenergic-specific Flp recombinase driver controlled by the promoter of dopamine β-hydroxylase (Dbh), the enzyme required for the final step in norepinephrine synthesis. En1^Dre (Plummer et al., 2016) expresses Dre recombinase in the embryonic midbrain, r0, and r1 (Figure 1A), and Pax7^Cre (Keller et al., 2004) expresses Cre in the alar plate (Figure 1A). The indicator allele, RC::RFLTG (Plummer et al., 2015), expresses tdTomato after Dre-mediated deletion of a rox-flanked transcriptional stop cassette and Flp-mediated deletion of a FRT-flanked stop cassette. EGFP is expressed following Dre-, Flp-, and Cre-mediated deletion of all stop cassettes and the loxP-flanked tdTomato.

In mice heterozygous for all three recombinase driver alleles and the indicator allele, noradrenergic neurons derived from the Pax7^Cre-defined alar plate of r0 and r1 (the intersectional population) will be selectively labeled with EGFP (Figure 1B). Any noradrenergic neurons derived from r0 and r1 outside of the Pax7^Cre domain (the subtractive population) will be labeled with tdTomato, and noradrenergic neurons originating outside of r0 and r1 will not be labeled due to the presence of the rox-flanked stop cassette (Figure 1B). Co-labeling of the same neuron with EGFP and tdTomato is not possible, because of the order in which driver expression occurs. Pax7^Cre activity can be detected at E10.5 (Figure 1A), prior to the detection of Dbh^Flpo activity (Robertson et al., 2013). Therefore, Pax7^Cre-mediated recombination will delete the tdTomato cassette in progenitor cells before Dbh^Flpo-mediated recombination in postmitotic noradrenergic neurons permits fluorescent protein expression.

In the developing hindbrain of E15.5 embryos heterozygous for all three recombinase drivers and the indicator allele, we observed predominantly EGFP-labeled Pax7^Cre-positive noradrenergic neurons (Figure 2), consistent with the prior observation that the locus coeruleus is derived from the alar plate (Aroca et al., 2006). However, we consistently observed a small population of tdTomato-labeled Pax7^Cre-negative noradrenergic neurons intermingled with the EGFP-labeled neurons (Figure 2). As expected, in control littermates heterozygous for En1^Dre and Dbh^Flpo, but not Pax7^Cre, all r0- and r1-derived noradrenergic neurons were labeled with tdTomato, and no labeling was observed in other genotypes. These results indicate that a subpopulation of r0- and r1-derived noradrenergic neurons originates outside the Pax7^Cre-defined alar plate.

To determine whether this subpopulation contributes to all three regions of the LC complex, we tracked r0- and r1-derived noradrenergic neurons into the adult brain (n = 10 mice) where the LC, SubC, and A7 noradrenergic nuclei can be readily identified. To confirm that fluorescent labeling was restricted to noradrenergic neurons, we immunolabeled with an antibody against tyrosine hydroxylase (TH). We consistently observed Pax7^Cre-negative noradrenergic neurons (tdTomato+, TH+) intermingled with Pax7^Cre-positive noradrenergic neurons (EGFP+, TH+) in all three regions of the LC complex (Figure 3). As expected, in Pax7^Cre-negative controls heterozygous for En1^Dre and Dbh^Flpo, we observed tdTomato in all r0- and r1-derived noradrenergic neurons, and no noradrenergic neurons were labeled in other controls. Thus, we have uncovered two distinct populations within the noradrenergic primordium of En1-defined r0 and r1 that contribute to the adult LC complex.

To fully visualize the three-dimensional distribution of these two populations of noradrenergic neurons, we used the passive clarity technique (Yang et al., 2014) to clear adult brain tissue. Imaging and three-dimensional rendering of the intact LC complex (n = 6 mice) demonstrated that Pax7^Cre-negative (tdTomato-labeled) and Pax7^Cre-positive (EGFP-labeled) noradrenergic neurons are intermingled throughout the rostrocaudal and dorsoventral axes (Figure 4A and Supplementary Video 1). Next, we performed counts of tdTomato and EGFP-labeled neurons (n = 6 mice) to determine the relative contribution of the two subpopulations (Figure 4B). In the bilateral LC complex, we counted 2,326 ± 36 labeled neurons, a value intermediate between two previous estimates of cell numbers in the LC, including an undefined portion of the SubC, from C57BL/6 mice (Berger et al., 1979; Touret et al., 1982). Of that total, 208 ± 18 neurons were Pax7^Cre-negative (tdTomato-labeled) and 2,118 ± 36 were Pax7^Cre-positive (eGFP-labeled). Thus, the Pax7^Cre-negative population constitutes 8.93 ± 0.74% of the full LC complex. Because the LC complex was imaged bilaterally, we were able to compare left and right sides. We observed no significant difference in the number of Pax7^Cre-negative (p = 0.755, unpaired t-test) or Pax7^Cre-positive neurons (p = 0.677, unpaired t-test).

To determine if the Pax7^Cre-negative population defines a morphologically distinct subset of LC complex neurons, we examined the tissue cleared by PACT. Studies of LC neuron morphology in the rat have previously indicated the existence of three main cell types: medium sized multipolar neurons, medium-sized fusiform neurons, and a few small ovoid neurons (Swanson, 1976; Shimizu et al., 1978; Cintra et al., 1982). Retrograde tracing and immunolabeling has demonstrated that both classes of medium-sized neurons are noradrenergic projection neurons (Loughlin et al., 1986). The identity of the small ovoid cells is uncertain, and different studies disagree as to whether they are noradrenergic (Shimizu et al., 1979; Loughlin et al., 1986). Because the tdTomato-labeled, Pax7^Cre-negative neurons were dispersed in the LC complex, we could assess the shape of individual cells, similar to previous studies utilizing Golgi staining (Shimizu et al., 1978; Cintra et al., 1982). Roughly one quarter of the neurons examined were fusiform, and three quarters were multipolar (Figure 5). We did not observe any neurons that could be unequivocally assigned to the small ovoid category. Thus, the subpopulation of LC complex neurons originating outside the Pax7^Cre domain appears to include both morphologies of medium sized noradrenergic projection neurons.

Despite its relatively small size, this Pax7^Cre-negative subpopulation is likely functionally significant if its axon projection profile differs from that of the larger Pax7^Cre-positive subpopulation. To begin to address this question, we compared the axon projection profiles of the two subpopulations by exploiting the unique ability of the RC::RFLTG indicator allele to label axons of both the Pax7^Cre-negative (tdTomato+) and Pax7^Cre-positive (EGFP+) populations. For the most sensitive detection of axons, we labeled Pax7^Cre-positive (EGFP+) and Pax7^Cre-negative (tdTomato+) axons separately in sections from adult brain (n = 8 mice) using a horse radish peroxidase-linked secondary antibody and 3,3′-diaminobenzidine (DAB) staining. Axons from the LC complex project throughout the brain and provide the major source of norepinephrine to the cortex and thalamus (Robertson et al., 2013). Upon examination of sections labeled for the Pax7^Cre-negative population, we observed a striking absence of axonal inputs to the thalamus (Figure 6). In contrast, inputs from the Pax7^Cre-positive population were consistently present in thalamus, and we observed axons from both populations throughout the cortex, although there were fewer Pax7^Cre-negative axons, as expected for a smaller population (Figure 6). This axonal projection pattern suggests that the Pax7^Cre-negative neurons may be functionally distinct from the larger population of Pax7^cre-positive LC complex neurons, highlighting the importance of early gene expression in generating differences in the neuronal circuitry.

Previous studies have clearly shown that neurons of the locus coeruleus are heterogeneous with respect to anteroposterior origin. Our initial intersectional fate mapping experiments demonstrated that the LC is primarily derived from the En1 expression domain encompassing r0 and r1, with a very small subpopulation derived from r2 (Robertson et al., 2013). The more recent single-gene fate mapping using Fgf8^Cre showed that noradrenergic neurons derived from r0, as defined by Fgf8 expression, contribute to the dorsal LC while those derived from r1 contribute to the ventral LC (Watson et al., 2017). Our results here reveal that the LC is also developmentally and genetically heterogeneous with respect to dorsoventral origin. Furthermore, the portions of the SubC and A7 nuclei included within the En1-derived LC complex are similarly composed of Pax7^Cre-positive and Pax7^Cre-negative subpopulations. The projection patterns of the two Pax7^Cre-defined subpopulations of the LC complex may point to functional differences; the virtual absence of projections to the thalamus suggests that the Pax7^Cre-negative subpopulation does not directly participate in the modulation of thalamic neuron activity performed by the broader LC complex (Berridge and Waterhouse, 2003). Taken together, these new data revealing (1) further heterogeneity within the anatomically defined LC, and (2) genetically defined subpopulations spanning several anatomically defined noradrenergic nuclei, offer further support for the idea that the anatomical divisions of the noradrenergic system are artificial constructs overlaid on natural genetic neuroanatomy arising from developmental history.

Limitations of the current study are the inability to overlay our Pax7^Cre-defined noradrenergic fate map on the recently described Fgf8-defined subpopulations of the LC (Watson et al., 2017), and uncertainty regarding the precise dorsoventral origin of the Pax7^Cre-negative subpopulation. We observe Pax7^Cre-negative noradrenergic neurons throughout the dorsoventral extent of the adult LC complex, suggesting that they originate in both r0 and r1. However, it is possible that they originate in one domain and then migrate among the neurons derived from the other. A subpopulation as small as the Pax7^Cre-negative noradrenergic neurons could easily be overlooked in single-gene fate-mapping experiments such as that which defined the dorsal r0-derived and ventral r1-derived LC subpopulations. Additional intersectional fate mapping experiments will be required to answer this question.

We are also unable to determine the precise dorsoventral level of the neural tube that is the source of the Pax7^Cre-negative subpopulation of the LC complex, but the basal plate is a likely candidate. Ascl1 (Mash1), a basic helix-loop-helix transcription factor required for noradrenergic neuron differentiation (Hirsch et al., 1998), is expressed in both alar and basal domains (Pattyn et al., 2004). It seems plausible that the basal Ascl1 expression domain could give rise to a small population of noradrenergic neurons. Other possibilities include the floor plate, known to be the origin of dopaminergic neurons in the En1-expressing mesencephalon (Kittappa et al., 2007; Ono et al., 2007; Bonilla et al., 2008; Joksimovic et al., 2009; Blaess et al., 2011), or an as yet undetected alar microdomain which does not express Pax7^Cre. We cannot fully rule out mosaic Cre expression within the alar Pax7 expression domain as the source of the Pax7^Cre-negative subpopulation, but the consistent size of the subpopulation and its distinct projection profile argue against this interpretation of the data. Efficient Pax7^Cre-mediated recombination of a fluorescent Cre indicator that is a direct derivative of RC::RFLTG (Figure 1A) also indicates that the Pax7^Cre-negative population is not an artifact of incomplete recombination. Regardless of origin, however, differential Pax7^Cre expression reveals two subpopulations of the LC complex that are molecularly distinct and experimentally separable.

Without a marker for the precise dorsoventral domain in which the Pax7^Cre-negative subpopulation arises, its function cannot be experimentally assessed using genetic tools that depend on recombinase expression to drive intersectional expression of effector molecules such as hM3Dq (Sciolino et al., 2016). However, our ability to fluorescently label these two LC complex subpopulations would permit transcriptional profiling to identify differentially expressed genes, and subsequent generation of a recombinase driver expressed in the Pax7^Cre-negative subpopulation. These future investigations promise to extend our knowledge of how embryonic cells with different developmental origins combine to give rise to the molecular and cellular heterogeneity of the mature LC complex. Such knowledge will be required for an understanding of how heterogeneity within the LC complex contributes to its regulation of diverse behaviors in the adult, and the roles that its dysfunction plays in neurological disease.

PJ and NP designed and planned the experiments. PJ, NP, ES, KS, and CT performed the experiments. PJ, NP, KS, and ES analyzed the data. NP and ES prepared the figures. PJ and NP wrote the manuscript with input from co-authors.