Genetic inducible fate mapping in mouse: Establishing genetic lineages and defining genetic neuroanatomy in the nervous system

Authors

  • Alexandra L. Joyner,

    Corresponding author
    1. Howard Hughes Medical Institute and Developmental Genetics Program, Skirball Institute of Biomolecular Medicine, Departments of Cell Biology and Physiology and Neuroscience, New York University School of Medicine, New York, New York
    • Howard Hughes Medical Institute and Developmental Genetics Program, Skirball Institute of Biomolecular Medicine, Departments of Cell Biology and Physiology and Neuroscience, New York University School of Medicine, 540 First Avenue, New York, NY 10016
    Search for more papers by this author
  • Mark Zervas

    1. Howard Hughes Medical Institute and Developmental Genetics Program, Skirball Institute of Biomolecular Medicine, Departments of Cell Biology and Physiology and Neuroscience, New York University School of Medicine, New York, New York
    Search for more papers by this author

Abstract

A fascinating aspect of developmental biology is how organs are assembled in three dimensions over time. Fundamental to understanding organogenesis is the ability to determine when and where specific cell types are generated, the lineage of each cell, and how cells move to reside in their final position. Numerous methods have been developed to mark and follow the fate of cells in various model organisms used by developmental biologists, but most are not readily applicable to mouse embryos in utero because they involve physical marking of cells through injection of tracers. The advent of sophisticated transgenic and gene targeting techniques, combined with the use of site-specific recombinases, has revolutionized fate mapping studies in mouse. Furthermore, using genetic fate mapping to mark cells has opened up the possibility of addressing fundamental questions that cannot be studied with traditional methods of fate mapping in other organisms. Specifically, genetic fate mapping allows both the relationship between embryonic gene expression and cell fate (genetic lineage) to be determined, as well as the link between gene expression domains and anatomy (genetic anatomy) to be established. In this review, we present the ever-evolving development of genetic fate mapping techniques in mouse, especially the recent advance of Genetic Inducible Fate Mapping. We then review recent studies in the nervous system (focusing on the anterior hindbrain) as well as in the limb and with adult stem cells to highlight fundamental developmental processes that can be discovered using genetic fate mapping approaches. We end with a look toward the future at a powerful new approach that combines genetic fate mapping with cellular phenotyping alleles to study cell morphology, physiology, and function using examples from the nervous system. Developmental Dynamics 235:2376–2385, 2006. © 2006 Wiley-Liss, Inc.

GENETIC FATE MAPPING IN MOUSE: AN OVERVIEW

Physical fate mapping approaches involve marking a cell, or group of cells, such that its descendents can be followed during development. Ideally, the technique allows specific cells to be marked based on their position, cell type, or genetic makeup. Because genetic fate mapping achieves these ideals and is noninvasive, it has an enormous advantage over physical methods for marking cells. In addition, a genetically marked population can be carefully characterized, whereas most other methods require retrospective analysis. The basic strategy underpinning genetic fate mapping (Fig. 1) uses two genetically engineered alleles in mice to mark and visualize individual cells over time (Zinyk et al., 1998). One allele expresses a site-specific recombinase, whereas the other is a Reporter allele that is engineered to permanently and heritably express a marker following site specific recombination. Two main recombinases have been used for fate mapping: Cre (derived from bacterial phage), which acts on loxP sites, and Flp (derived from S. cervisiae), which acts on frt sites (for reviews, see Dymecki, 2000; Branda and Dymecki, 2004). Depending on the purpose of the fate mapping study, Cre/Flp are expressed in a gene-specific manner, allowing an individual cell type, distinct genetic lineage, or a particular spatial domain of a tissue to be indelibly marked. This level of spatial control is achieved by choosing an appropriate gene regulatory element to express Cre/Flp and making standard transgenic mice (e.g., Zinyk et al., 1998; Dymecki and Tomasiewicz, 1998) or knockin alleles by gene targeting (e.g., Michael et al., 1999; Kimmel et al., 2000). Several conditional Reporter alleles have been generated that contain a loxP- or frt-flanked STOP cassette located upstream of various markers under the control of a promoter that is broadly expressed using transgenes or knockin alleles (e.g., lacZ, Soriano, 1999, or Rodriguez and Dymecki, 2000; alkaline phosphatase, Lobe et al., 1999, or Awatramani et al., 2001; enhanced green fluorescent protein [eGFP], Novak et al., 2000, or Awatramani et al., 2003). For example, in a mouse carrying the R26R reporter allele (Soriano, 1999) and a Cre allele, the loxP-flanked STOP sequence is deleted specifically in cells that express Cre, and these cells and all their descendents are permanently marked by expression of lacZ. The marking generated by Cre/Flp occurs in all of the cells that have ever expressed Cre/Flp and, thus, is cumulative. Finally, by combining Cre- and Flp-mediated recombination with a single reporter allele that has both loxP- and frt-flanked STOP cassettes, a more precisely defined population of cells can be marked, in an approach that allows for intersectional and subtractive fate mapping (Awatramani et al., 2003; Farago et al., 2006). The type of cell that can be marked by genetic fate mapping is only limited by the expression of each specific reporter allele and needs to be determined empirically to evaluate the full extent of putative cell lineages that can be marked with each reporter. The R26R reporter has thus far proven to be an ideal reporter because of its expression in embryonic and adult cells, including those of the central and peripheral nervous systems, limb, skin, lung, and heart.

Figure 1.

Genetic inducible fate mapping strategy. Schematic illustrations of the genetic components and cellular events underpinning temporal control of recombination. A: The primary components of Genetic Inducible Fate Mapping (GIFM) are shown at the level of the cell (left illustration). CreER (green oval) is sequestered in the cytoplasm by heat shock protein 90 (Hsp 90; orange oval). The R26R (rosa) reporter allele consists of a stop sequence (red line) flanked by two loxP sites (triangles). The floxed stop cassette precedes the lacZ gene, which although present in all cell types, is quiescent in R2R6 reporter mice. CreER expression is controlled by placing it under the control of region- or cell type-specific genes; in this example, CreER is expressed in the mouse Hb (right illustration, gray ovals). B: After the administration of tamoxifen (red T), it binds to the CreER fusion protein, which translocates to the nucleus and seeks out loxP sites (beginning approximately 6 hr after tamoxifen administration). C: CreER induces recombination between loxP sites, which removes the stop cassette and allows the lacZ gene to be turned on; tamoxifen pharmacokinetics confers that the recombination event will occur for only an additional 24–36 hr (Robinson et al., 1991). D: Once lacZ is expressed, it begins to produce β-galactosidase, which can be detected by X-gal (left illustration, blue). Critical to GIFM is that the recombined allele is constitutive and heritable. The temporal and spatial control confers marking of a small population of cells in the Hb (right illustration); the contribution of the marked cells can then be followed (arrows).

A significant advance in genetic fate mapping has been the introduction of inducible forms of Cre in mice to accurately control when cells are marked (Brocard et al., 1997; Kimmel et al., 2000; Zirlinger et al., 2002; Guo et al., 2003; Ahn and Joyner, 2004; Harfe et al., 2004; Zervas et al., 2004). We have termed such an approach Genetic Inducible Fate Mapping, or GIFM (Fig. 1). The most common strategy for temporally regulating Cre activity has been to generate Cre fusion proteins with a tamoxifen-responsive Estrogen Receptor ligand binding domain from the human (ERT, Feil et al., 1996 and Brocard et al., 1997; ERT2, Feil et al., 1997) or mouse (ERTM, Hayashi and McMahon, 2002) gene. CreER is sequestered in the cytoplasm by heat shock protein 90 (Hsp 90). The administration of tamoxifen confers a conformational change on CreER that releases it from Hsp 90 and allows it to be translocated to the nucleus, where it induces recombination between loxP sites (Fig. 1A–C). The recombination event that activates the reporter allele occurs within 6–12 hr and continues for up to 36 hr (e.g., Hayashi and McMahon, 2002; Zervas et al., 2004; and unpublished results), which is consistent with tamoxifen pharmacokinetics in vivo (Robinson et al., 1991). A FlpERT2 fusion protein also was shown recently to induce recombination ∼18 hr after administration of tamoxifen (Hunter et al., 2005), although the level of recombinase activity appears to be lower than with CreER (unpublished results). The initial marked population is then determined 24–48 hr after administration of tamoxifen by detection of the marker protein, to ensure that the marked cells recapitulate the expected gene expression. Because marking with GIFM is highly reproducible, the initially marked population can be further characterized by comparing labeled cells to morphological landmarks and marker genes. The fate of the initial population is then visualized at any later time point. In this way, the long-term fate and contribution of cells not only from a distinct genetic lineage but importantly from cells marked during a defined 24-hr period are ascertained. Labeling cells only at a precise time circumvents limitations associated with cumulative fate mapping using Cre/Flp, such as the masking of discreet subpopulations in a dynamically changing gene expression domain. The most important feature of inducing Cre activity at different time points is that temporal aspects of lineage allocation can be addressed. The robustness and speed of marking depends on the ER component and the level of expression of the CreER fusion protein. This dependence may represent a limitation when looking at genes expressed in a gradient because putatively marked cells in a low-expression domain may appear as under-represented in comparison to those of a higher expression domain.

One way to address the issue of cells that express genes at low levels is to give relatively high doses of tamoxifen to ensure more robust marking. In our hands, there is a quantitative increase in marking using the R26 reporter allele when giving relatively low (2.0 mg/40 g mouse) versus intermediate (4.0 mg/40 g mouse) or high (10.0 mg/40 g mouse) doses of tamoxifen. However, there does not appear to be a linear relationship between tamoxifen dose and recombination efficiency. An additional method to increase recombination efficiency is to use a second generation inducible Cre (CreERT2), which has been shown to be between 4- and 10-fold more sensitive to 4-hydroxytamoxifen in vivo (Feil et al., 1997; Indra et al., 1999). 4-hydroxytamoxifen is a metabolite of tamoxifen that is generated in the liver and believed to be the active compound in binding the ER, although administration of either compound, at the same concentration, results in indistinguishable recombination efficiency (Kuhbandner et al., 2000). Importantly, embryonic stages are particularly sensitive to high tamoxifen doses and may result in lethality, especially in inbred strains. Again, in our experience, intermediate doses are generally well tolerated in Swiss Webster female mice, giving efficient recombination and good litter viability. Oral gavage delivery of tamoxifen appears to reduce litter loss apparently associated with intraperitoneal inflammation resulting from direct injection (unpublished observation), with both methods providing similar recombination efficiency (Kuhbandner et al., 2000). Progesterone (approximately 1.5–2.0 mg/40 g pregnant female) also appears to improve litter viability when giving high doses of tamoxifen (unpublished observation). Clearly, all of these issues and factors need to be considered and addressed based on individual experimental conditions.

In a recent extension of GIFM, the level of tamoxifen was lowered such that single cells were marked, allowing for clonal analysis in the skin (Legue and Nicolas, 2005). An alternative approach for clonal analysis was adapted recently to mouse from fruit fly. The MADM (mosaic analysis with double markers) technique relies on rare Cre- or CreER-driven recombination events between two homologous chromosomes to activate a different marker gene on each chromosome (Zong et al., 2005). If recombination occurs during G2, then X segregation of the chromosomes (production of two recombinant sister chromatids that segregate into different daughter cells) will result in each cell expressing one or the other marker. Collectively, the genetic fate mapping studies to date likely illustrate only the tip of the iceberg of emerging technologies that will be developed in combination with GIFM (see below for future directions). Furthermore, while hundreds of Cre transgenic and targeted mouse lines have been produced (see the evolving creX mice database at http://nagy.mshri.on.ca/PubLinks/indexmain.php), only a small number of CreER and FlpER lines have been published. Thus, GIFM only recently has begun to gain momentum, with the greatest concentration of papers focusing on the nervous system and in particular the anterior hindbrain.

Temporal and Spatial Partitioning of Neurons in the Anterior Hindbrain

Genetic fate mapping has uncovered new principles and confirmed previous hypotheses of how the cerebellum (Cb) develops. In addition, these methods have revealed how nuclei of the anterior hindbrain (Hb) are generated and organized based on their position, genetic lineage, and time of emergence from the dorsal neural tube. GIFM has been essential in determining the genetic and cellular events that control anterior Hb development because of the methods ability to define when and from where each of the different cell types arises.

Development of the Anterior Hindbrain

The mouse nervous system develops from the dorsal ectoderm through the bending and zipping up of the neural plate to form a neural tube. At embryonic day (E) 8.5, the developing brain is clearly visible as bilateral head folds. Cells at the lateral edges of the neural plate migrate ventrally to form the peripheral nervous system and the bones of the head. By E9.5, the neural tube has formed and is divided morphologically along the anterior–posterior (A-P) axis (see Figs. 1A, 2A). The segmentation roughly translates into the broad functional domains of the adult central nervous system: forebrain, midbrain (Mb), Hb, and spinal cord. The neural tube is initially a single cell layer of proliferating neural progenitors, called the ventricular zone (VZ), which lines the ventricles and harbors a resident population of cells that undergo asymmetric divisions to produce both postmitotic neurons and additional neural progenitors. Starting at ∼E9.5, progenitors begin to produce postmitotic cells that migrate out and either integrate into layered brain structures (for example, in the dorsal Mb and Cb), or condense into cell groups anatomically arranged into clusters referred to as nuclei (for example, dopaminergic and serotonergic neurons of the ventral Mb and Hb, respectively). Genetic fate mapping has proven to be a powerful tool for delineating precise migratory paths of specific neurons and for unraveling how they become organized into these layers or nuclei (see below).

Figure 2.

Genetic inducible fate mapping summary. Genetic Inducible Fate Mapping (GIFM) of Wnt1-, Math1-, and En1-derived lineages marked by tamoxifen administration and CreER-mediated recombination of the R26R reporter in the mid/hindbrain in vivo. A: Schematic of embryonic day (E) 12.5 embryo showing the distribution of early marked populations when tamoxifen was administered in utero at E10.5. Wnt1-CreER marks cells in the midbrain (Mb, dark blue circles) and in the lower rhombic lip (LRL, light blue circles) of the hindbrain (Hb); Math1-CreER marks cells of the upper RL (URL, dark red circles) and in the LRL (light red circles). A small domain of cells marked using En1-CreER is confined to the medial region of r1 (note: En1-CreER marked cells, which are also present in the Mb, are illustrated only in the Cb). Wnt1-derived cells of the posterior midbrain expand only anteriorly and do not cross a lineage boundary (1, dark blue arrow). Wnt1-derived cells of the LRL contribute to the medullary velum overlying the 4th ventricle (2, light blue arrow), which later becomes the choroid plexus; Wnt1-derived cells of the LRL also stream ventrally in the Hb (3, light blue arrows). In contrast, Math1-derived cells of the LRL do not contribute to the choroid plexus, but do stream ventrally (4, red arrows). A′: Summary of some of the important gene expression domains (left plane) of the LRL ventricular zone (vz): Gdf7 is expressed in the dorsal roof plate (purple), Math1 is expressed in a more ventral–lateral domain (yellow), and Wnt1 is expressed in an apparently subtle dorsal to ventral gradient (blue–white). Two Wnt1-Flp transgenes illustrate a further partitioning of the LRL (Landsberg, 2005; see text for details). Note: only part of the Wnt1 expression domain is shown superimposed on the Math1 and GDF7 domains in the left illustration but is shown to be expressed throughout the domain as illustrated on the right. Comparative fate mapping schematic (right plane) showing that Wnt1-CreER marks a resident pool of Wnt1-derived precursors (Zervas et al., 2004), whereas a Math1-CreER transgene marks differentiating cells (Machold, 2005). These cells travel ventrally en route to Hb destinations. B: Schematic of the fate of cells marked by tamoxifen at E10.5 at an intermediate time point (E14.5). Midbrain Wnt1-derived cells (blue circles) expand rostrally and remain in the posterior Mb, whereas URL Math1-derivatives are located in an anterior dorsal position in the cerebellum (Cb, see cut away to show sagittal view) and begin to migrate ventrally (6, red arrow). In contrast, r1 Math1-derivatives marked with tamoxifen at E12.5 (black circles) migrate over the E14.5 Cb (5, black arrows). LRL Wnt1 and Math1 derivatives marked at E10.5 are in transition (7, 8) to their final position in the Hb. C: The final destination of marked cells in the adult. Wnt1-derived cells of the Mb (blue circles) reside in the inferior colliculus (ic), whereas medially derived En1-derived cells in the Cb (green circles) contribute to all cell types of the vermis (see also E). Math1-expressing URL cells marked at E10.5 contribute to the deep cerebellar nuclei (dcn) and anterior Hb nuclei (red circles) and cells marked at E12.5 to the granular layer (gl, black circles). LRL Wnt1 derivatives marked at E10.5 give rise to the pontine (pn), vestibular (vn), and cuneate nuclei (cn), whereas Math1 derivatives contribute to the superior olive (so) as well as the vn, and cn. D: GIFM also has demonstrated how dorsal r1 undergoes an orthogonal rotation such that anterior r1 (gray) becomes medial Cb, whereas progressively more posterior r1 domains (orange and yellow, respectively) become laterally organized domains in the adult Cb; this is true of all but granule cells (see text for details; adapted from Sgaier et al., 2005). E: Distribution of marked populations within the adult Cb. En1 derivatives (green circles) contribute to all layers of the Cb including the molecular layer (ml), Purkinje cell layer (P), and gl. Early marked Math1-expressing cells (red circles) contribute only to the dcn, whereas later derivatives give rise only to the gl (black circles).

The anterior Hb has several developmental properties that collectively make this a unique region of the central nervous system. The Hb is divided along the A-P axis into seven compartments in mouse called rhombomeres (r1–r7) that are apparent morphologically as bulges, express distinct genes, and are separated by lineage boundaries. The Hb neural tube is not closed dorsally during early embryogenesis causing most of dorsal r1 to splay open with a thin membrane covering the 4th ventricle (Fig. 2A). Thus, dorsal r1 appears as two bilateral wing-like structures that are of considerable distance from each other. Dorsal r1 forms the Cb, which is a laminated structure composed of an outer molecular layer interspersed with interneurons, an adjacent layer of Purkinje cells, followed by a broad layer of granule cells, and finally a deep core that houses the deep cerebellar nuclei (Fig. 2C,E). In contrast, ventral r1 and the remaining rhombomeres generate the Hb that is largely organized into precisely positioned nuclei (Fig. 2C). The dorsal neural precursors that abut the membrane covering the 4th ventricle during embryogenesis form a specialized germantive zone called the rhombic lip (RL). The RL is divided into the upper RL (URL) associated with r1 and the lower RL (LRL) associated with more posterior rhombomeres, based on the characteristics of the neurons they each generate. The neurons born in the LRL have the special property of migrating long distances (ventrally) away from their site of origin (Fig. 2A,B; arrows 3,4,7,8). Genetic fate mapping, in particular inducible and intersectional approaches, has proven to be extremely valuable in determining when these neurons are formed and from which RL domains certain ventral hindbrain nuclei and the Cb are produced (reviewed below). In general, cells generated from the URL migrate anteriorly over the surface of r1 (Fig. 2B, arrow 5). Like cells of the LRL, the early born cells of the URL ultimately migrate ventrally to contribute to specific hindbrain nuclei, as well as the deep cerebellar nuclei (Fig. 2B, arrow 6). In contrast, granule cell precursors are generated later (after ∼E13.5) from the URL. They remain on the surface of the developing Cb and continue to proliferate postnatally (Fig. 2B, arrow 5). When they become postmitotic, they migrate into the depths of the Cb. Recent fate mapping studies have begun to define when and where each cell type is generated, as well as the spatial coordinates of the Cb (reviewed below).

Genetic Fate Mapping of the Anterior Hindbrain

Genetic fate mapping in mouse has clarified seemingly incongruent fate mapping results in chick regarding the presumptive origin of the Cb. Early transplantation studies in chick relying on morphological landmarks to mark cells proposed that the anterior–medial Cb (except for granule cells) arises from the posterior Mb. However, a later study using quail–chick chimeras using the caudal expression of Otx2 as a landmark of the posterior Mb proposed that the Cb arises from the anterior Hb (reviewed in Zervas et al., 2005). It was shown using GIFM and regulatory sequences from Wnt1, which in the Mb has the same posterior expression limit as Otx2, that the cerebellum arises exclusively from the anterior Hb in mouse. In contrast, a small population of cells located just anterior to the Mb/Hb lineage boundary expands rostrally and contributes to the posterior Mb (Zervas et al., 2004). By comparing GIFM of the Mb to cumulatively marked cells using additional Cre lines, it was further demonstrated that a series of lineage restriction boundaries partition the developing Mb and Cb into distinct compartments (Zervas et al., 2004). In a subsequent study, it was shown that, at E9.5, the neural tube undergoes a 90-degree rotation such that the A-P axis of the neural tube of r1 is converted into the medial–lateral axis of the adult Cb (Fig. 2D; Sgaier et al., 2005). In this way, GIFM delineated morphogenetic movements that had been suggested from transplantation studies in chick and random genetic marking of clones in mouse (Hallonet and Le Douarin, 1993; Alvarez Otero et al., 1996; Mathis et al., 1997). In additional experiments, distinct medial–lateral regions of the cerebellar primordium at E12.5 were marked and their fate determined by inducing marking at different time points using two En-CreER lines (Sgaier et al., 2005). These studies showed that all but the granule cells maintain their initial general position through to the adult, despite the complex morphogenesis involved in generating folia in the Cb (Fig. 2D). The granule cells, however, undergo an unexpected lateral to medial migration to populate the last folds that form in the posterior Cb. Finally, MADM was used to study granule cell precursor clones and uncovered that descendents of these clones marked at E17.5 all project their axons to a particular sublayer of the cerebellar molecular layer (Zong et al., 2005).

Physical fate mapping studies in chick showed that the URL is the source of granule cells and unexpectedly indicated that some URL-derived cells migrate ventrally to form precerebellar nuclei in the isthmus (junction region between the Mb and Hb) and ventral r1 (Marin and Puelles, 1995; Ambrosiani et al., 1996; Wingate and Hatten, 1999; Lin et al., 2001). GIFM and cumulative marking of r1 in mouse confirmed this unexpected mode of migration and uncovered that additional nuclei, including the deep cerebellar nuclei, arise from the URL. GIFM also showed that the URL and VZ of r1 is further subdivided into discrete subpopulations based on gene expression. In one study, a Math1 transgene was used to express CreER in the entire RL and dorsal midline of the nervous system. By comparing GIFM to En1-Cre cumulative fate mapping (Li et al., 2002; Zervas et al., 2004), the r1-derived cells were more precisely identified (Machold and Fishell, 2005). Furthermore, by marking Math1-CreER–expressing cells at different time points using tamoxifen, the temporal sequence of generating URL-derived ventral nuclei, the deep cerebellar nuclei, and the granule cells was determined (Table 1; Machold and Fishell, 2005). The same cell types were labeled in Math1-lacZ knockin mice using perdurance of LacZ protein as a short-term cumulative marker as well with Math1-Cre cumulative fate mapping (Table 1; Landsberg et al., 2005; Wang et al., 2005; and see Wingate, 2005). Importantly, these studies illustrate that the various nuclei derived from the URL of r1 are precisely delineated early, based on the timing of expression of one gene. These allocations of the URL would not be possible using traditional cell marking methods. Of interest, the VZ of the cerebellum primordia (r1) also can be separated into at least two genetically separable subpopulations: γ-aminobutyric acid (GABA)ergic neuron precursors that transiently express Ptf1 (using a Ptf1-Cre knockin allele; Kawaguchi et al., 2002), and glutamatergic neuron precursors that do not. Of interest, although GABAergic interneurons in the deep nuclei are marked with Ptf1-Cre, some deep cerebellar neurons are marked with Math1-CreER, demonstrating that the two cell types are born in distinct locations and then migrate and coalesce to form a particular deep nucleus.

Table 1. Summary of GIFM and Cumulative Genetic Fate Mapping of the Hb Showing Structures Are Derived From Cells of Distinct Genetic Lineagesa
Hindbrain structures labeledWnt1-CreER markingMath1-CreER markingX-Cre markingb
E9.5E10.5E10.5E12.5Cumulative
  • a

    GIFM of Wnt1-CreER–marked cells (Zervas et al., 2004). GIFM of Math1-CreER–marked cells (Machold and Fishell, 2005). GIFM, Genetic Inducible Fate Maping; Hb, hindbrain; E, embryonic day; URL, upper rhombic lip; nd, not determined; LRL, lower rhombic lip; n., nuclei; ✓, contribution; —, no contribution.

  • b

    X-Cre marking indicates cumulative marking approaches with references as follows: 1, Wnt1-Flp (Rodriguiz and Dymecki, 2000; Awatramani et al., 2003; Landsberg et al., 2005); 2, Gdf7-Cre (Currle et al., 2005); 3, En1-Cre (Zervas et al., 2004; Machold and Fishell, 2005); and 4, Math1-LacZ (Wang et al., 2005; lacZ perdurance as a cumulative approach) and Math1-Cre (Landsberg et al., 2005).

  • c

    Only a small number of positive granule cells was observed in posterior Cb.

Choroid plexus√(1, 2, 3)
Anterior Hb nuclei (URL-derived):     
 Raphe n.ndnd√(3)
 Parabigeminal n.√(1, 3, 4)
 Superior olive√(3, 4)
 Pedunculopontine n.√(1, 3, 4)
 Laterodorsal tegmental n.√(1, 3, 4)
 Microcellular tegmental n.√(1, 3, 4)
 Dorsal lateral lemniscus n.√(1, 3, 4)
 Lateral parabrachial n.√(1, 3, 4)
Pre-cerebellar nuclei (LRL-derived):     
 Pontine n.√(1, 4)
 Cuneate n.√(1, 4)
 Vestibular n.√(1, 4)
 Reticulotegmental n.ndndndnd√(1, 4)
 Lateral reticular n.ndndndnd√(1, 4)
 Inferior olivendnd√(1)
Cerebellum (r1/URL-derived):     
 Deep cerebellar nuclei√(3, 4)
 Purkinje cells√(3)
 Granule cellsc√(1, 3)
 Molecular layer√(3)

The LRL has also been genetically fate mapped along the dorsal–ventral (D-V) and A-P axes using several transgenes and knockins (see Pearse and Tabin, 2006). An interesting result is that the LRL appears to be divided into D-V gene expression domains that give rise to different cell types (Fig. 2A,A′; Landsberg et al., 2005). First, a gradient of Wnt1 was defined as spanning across the D–V domain (Fig. 2A′, blue–white gradient). By taking advantage of the Wnt1 gradient and the fact that one Flp transgene (FlpL) marks a smaller region than a second Flp transgene (Flpe), the LRL can be fate mapped into two broad domains (Landsberg et al., 2005). Furthermore, the Wnt1-Flp marked domain appears to correspond to most of the Math1-CreER/-lacZ/-Cre domain (Machold and Fishell, 2005; Wang et al., 2005; Landsberg et al., 2005). The LRL can be further partitioned into a most dorsal progenitor domain that expresses Gdf7 (Fig. 2A′). Accordingly, Gdf7-Cre marks cells closest to the epithelium covering the 4th ventricle, and exclusively gives rise to the choroid plexus (Fig. 2A,A′; Currle et al., 2005; Landsberg et al., 2005). Of interest, the broad dorsal domain marked by Math1-CreER/-Cre/-lacZ and Wnt1-Flp/-CreER gives rise to precerebellar nuclei in the ventral Hb that project mossy fibers to the granule cells, whereas the broader Wnt1-Flpe domain also marks the inferior olive, which houses the climbing fiber neurons that project to the Purkinje cells (Rodriguez and Dymecki, 2000; Zervas et al., 2004; Landsberg et al., 2005; Machold and Fishell, 2005; Wang et al., 2005). GIFM with Wnt1-CreER showed that the temporal contribution of the Wnt1 lineage to precerebellar nuclei occurs beginning at E10.5 (Fig. 2A–C; Table 1; Zervas et al., 2004). In addition, the choroid plexus is derived from the Wnt1 lineage, but before the contribution to the precerebellar system (Fig. 2A,A′; Zervas et al., 2004). Of interest, a GABA(A) receptor α6-Cre transgene marks a similar set of cerebellar neurons and precerebellar neurons (Funfschilling and Reichardt, 2002) to those that are marked by Math1-CreER/-lacZ/-Cre. Taken together, these studies show that the LRL cells that contribute to the two different major input systems of the Cb in the Hb are derived from distinct genetic lineages and domains. Moreover, because additional mossy fiber neurons are derived from the URL and spinal cord, if the Math1-expressing cells in the spinal cord also give rise to mossy fiber neurons, then all the mossy fiber input neurons and their granule cell targets are derived from the Math1 genetic lineage. Intersectional fate mapping, for example, combining the Wnt1-Flp transgenes with rhombomere-specific Cre lines of mice, has proven to be effective in defining which ventral hindbrain nuclei are derived from specific rhombomeres along the A-P axis (Awatramani et al., 2003; Farago et al., 2006). By coupling intersectional fate mapping to GIFM, it will be possible to achieve accurate positional information, temporal precision, and genetic history.

Diversified Uses of GIFM

In addition to defining genetic neuroanatomy and lineage relationships, genetic fate mapping offers a new approach for addressing many additional developmental questions. For example, the biology of cells sending and receiving growth factor signals can be studied from a new perspective using GIFM. In one set of studies in the limb, GIFM was used to follow the fate of cells expressing the morphogen Sonic hedgehog (Shh-CreER) from the zone of polarizing activity (ZPA), the organizing center that controls A-P patterning in the limb (Harfe et al., 2004). In a complementary study, the cells responding to Shh signaling (Gli1-CreER, a transcriptional target of Shh) were also fate mapped (Ahn and Joyner, 2004). One surprising result was that the ZPA itself, which is only a small group of Shh-expressing cells in the posterior limb, contributes to the three posterior digits. Furthermore, the earliest cells to leave the ZPA form digit 3, and cells leaving the ZPA progressively later contribute to digit 4 and finally digit 5. Of interest, whereas the cells that respond to Shh populate a larger area of the limb, they do not contribute to digit 1. Of interest, as development progresses, many of the cells that express Shh stop responding to Shh signaling. Thus, the position of a cell relative to the source of Shh alone does not determine the length or time a cell receives Shh signaling, disputing a simple morphogen model of Shh action (see Zeller, 2004).

Another area of in vivo biology that is not restricted to the nervous system and where GIFM is proving to be very powerful is in understanding stem cells. In most organs, there are quiescent stem cells that rarely divide; when they do, they give rise to transient amplifying cells that are fast dividing but short lived and produce one or multiple cell types. It has been difficult to mark the quiescent stem cells unambiguously, and cumbersome using retroviruses or dye marking. Several recent studies have expressed CreER in quiescent stem cells of the adult skin, hematopoietic system, and brain to determine the normal cell types derived from the different stem cell populations in vivo during normal homeostasis and after injury (Morris et al., 2004; Ahn and Joyner, 2005; Legue and Nicolas, 2005; Levy et al., 2005; Gothert et al., 2005). Except in the hematopoietic system, stem cells were found normally to give rise to a more restricted range of cell types than can easily be induced in vitro. In the adult brain, after marking the initial population of stem cells using GIFM, the rapidly dividing transient amplifying cells were killed with the mitotic inhibitor AraC. Using this paradigm, the quiescent stem cells were spared and followed over long periods. It was demonstrated that the stem cells continue to repopulate particular adult structures for at least a year, but they differentiate mainly into interneurons (Ahn and Joyner, 2005). Furthermore, by inducing Cre activity at different time points, it was determined at which stage the stem cell niche is established.

Future Prospects for Using GIFM to Analyze Cell Morphology, Physiology, and Function

The techniques that have been developed for GIFM provide the foundation for an exciting new way to study cell morphology, physiology, and function in mouse. With this novel approach, conditional alleles are being generated by the mouse genetics community to express proteins that allow an array of cellular attributes to be characterized, including detailed analysis of cell morphology, intracellular functions such as protein trafficking and signaling, cellular physiology, and the requirement for a particular cell type in development or homeostasis (Fig. 3A–D). We refer to these alleles as cellular phenotyping alleles. Several fluorescent proteins engineered to target them to particular cell compartments have been used to visualize cell shape and subcellular structure (reviewed in Hadjantonakis et al., 2003). The expression of eGFP in neurons has proven to be a very effective method for visualizing the morphology and projections of axons and dendrites, and, therefore, can be used to characterize local circuits, axon pathfinding, and axon innervation of target sites (Fig. 3B). To begin to map out more complete and/or functional circuits, trans-synaptic tracers are being developed. Building on a classic neuroscience tracing method using the plant lectin Wheat Germ Agglutinin (WGA), a cellular phenotyping allele has been generated with a transgene that conditionally expresses WGA, and the Cre activated WGA was shown to cross at least two synapses in an anterior grade manner (Fig. 3C; Braz et al., 2002). By combining GIFM and a cellular phenotyping allele that expresses eGFP and WGA, both local and long-range functional circuitry of neurons that are defined by genetic lineage can be studied (Fig. 3C). Defining the connection between genetic lineage and circuitry will likely be critical for a full understanding of how nervous system function is established as a result of early patterning events. In contrast to WGA, a nontoxic fragment of tetanus toxin (TTC) has been used for retrograde circuit mapping, and crosses at least one synapse (e.g., Maskos et al., 2002). The main limitation in these circuit mapping approaches is the sensitivity of visualizing the tracer and/or the distance it will travel, which seems to be correlated with the level of tracer expressed in each cell, It may be possible to boost the signal using self-enhancing genetic systems such as tTA/tetO. Another aspect of visualizing cells in vivo will be time-lapse imaging of cells to follow genetically defined cells in motion during development. In the developing mouse nervous system, this will be limited to in vitro culture preparations or in situ whole embryo cultures. The advent of genetic magnetic resonance imaging contrast agents could allow fate mapped cells to be tracked in vivo for many days (Cohen et al., 2005; Genove et al., 2005; Deans et al., 2006).

Figure 3.

Genetic Inducible Fate Mapping (GIFM) and cellular phenotyping. GIFM can be used with a diverse array of conditional alleles that are designed to assess cellular morphology, physiological properties, and function in the context of a normal working organ. We refer to the conditional alleles that allow functional analysis as cellular phenotyping alleles. A: The R26R reporter allele (Soriano, 1999) uses floxed stop lacZ and has been used widely in cell marking and fate mapping experiments (see text); Primary benefitgenetic lineage experiments. B: The use of eGFP to conditionally express green fluorescent protein (GFP) that fully labels neurons provides not only lineage information but also detailed neuronal morphology and the ability to trace axons as they pathfind and innervate their target areas and target neurons; Primary benefitlinking lineage to morphology. C: Functional connections can be determined using a conditional wheat germ agglutinin (WGA) WGA-eGFP allele (see Farago, 2003, for the generation of this allele). With this technique, first-order neurons of a distinct lineage will be marked by virtue of both WGA and eGFP expression (1). WGA is a trans-synaptic marker and will cross at functional synapses (a) into second-order neurons (2), an event that has the potential to occur at an additional synapse (b) thus marking a third-order neuron (3); Primary benefitgenetic lineage and functional circuits can be established. D: A conditional allele of diphtheria toxin (DT) is quiescent in neurons, but recombination-mediated removal of the stop cassette confers toxin accumulation in neurons (orange) and kills the cells due to the expression of the toxin (red skulls). A conditional DT receptor achieves similar results but has an added level of temporal control of when cells are ablated by requiring DT administration (see text for details); Primary benefitneurons of distinct lineages can be killed at select time points mimicking neurologic disorders. E: By using a region- or cell type-specific CreER (in this example, a cell type expressed in the Hb) and the various cellular phenotyping alleles can be expressed to maximize the potential of GIFM in vivo.

Another important extension of genetic fate mapping technologies will be to assess cell physiology and function. For example, to study physiology it should become feasible to express the array of proteins currently used in cells in culture to study calcium dynamics, intracellular trafficking, and ligand activated signaling pathways. The main stumbling block will be achieving a high enough level of expression in an inducible manner. To study the requirement for a certain cell type, the combination of GIFM and cellular phenotyping alleles harboring LoxP-STOP-diphtheria toxic (DT) A fragment or its receptor (e.g., Buch et al., 2005; Ivanova et al., 2005) will allow for the ablation of a genetically defined population of cells (Fig. 3D). The DT receptor is particularly powerful because it is neutral until DT is administered to the animal, thus providing a second level of temporal control (in addition to tamoxifen). The conditional removal of distinct neuronal lineages will be advantageous in modeling and understanding human neurological disorders using mice as a model. Finally, to determine the impact a neural cell has on animal behavior, neural activity blockers could be expressed from cellular phenotyping alleles (Yu et al., 2004). For these alleles, it will be necessary to express the blockers in an inducible manner from the population of cells marked with GIFM. One approach that has been developed is activation of tTA or rtTA from a loxP-STOP allele (Park et al., 2004; Belteki et al., 2005; Yu et al., 2005). The neural activity blocker would then be expressed from a tetO phenotyping responder allele.

A remaining challenge to those studying gene function in mice is a universal method for mutating a gene and marking the cell that carries the mutation. The MADM approach lends itself to marking and mutating rare cells. If one of the two marked chromosomes is linked to a mutant gene, then the daughter cell that expresses the reconstituted linked marker gene following Cre-mediated chromosomal exchange and X segregation will harbor a homozygous mutation. The limitation of this approach is not only the restriction to rare mutant cells but, more problematic, is the necessity that each mutant gene must be linked to a marker gene locus. In an attempt to mark cells and introduce a mutation using GIFM, we and others have used mice with both a floxed reporter allele (R26R) and a conditional gene inactivation allele. However, this approach can be misleading, because there can be a significant difference in the recombination efficiency of each allele such that the reporter allele recombines much more efficiently than the mutant allele (unpublished observation). In one experiment, Wnt1-CreER;R26R embryos carrying a floxed β-catenin allele were exposed to tamoxifen at multiple embryonic stages and analyzed for recombination of both alleles. Whole-mount labeling demonstrated a high degree of lacZ-marked cells in the developing Mb, although no mutant phenotype was observed. Consistent with the lack of a phenotype, the presence of recombined β-catenin alleles could not be detected using polymerase chain reaction (PCR, unpublished observation). Thus, in any given cell, the conditional mutant floxed allele, the marker allele, or both can recombine with varying efficiencies.

CONCLUSION

In summary, genetic fate mapping in mouse has proven to be a substantial improvement over physical labeling methods. Of the available genetic fate mapping approaches, GIFM is the most recent and significant advancement because of its diversity, as well as spatial and importantly temporal control of cell marking. In addition, the ability to indelibly mark cells of a distinct genetic lineage and follow them through development allows a link to be forged between early transient patterning events, cell behaviors, and the long-term contribution of small populations of marked cells to mature structures. Genetic fate mapping is now being extended by the introduction of new cellular phenotyping alleles. In the nervous system, this technique will be used to study how gene expression and lineage relate to cell morphology, axonal pathfinding, and target acquisition, as well as neuronal physiology and functional connectivity in vivo. An exciting future is at hand as GIFM and cellular phenotpying alleles are used in combination with gene inactivation to begin to understand brain function and neurological diseases using the mouse as a model organism. Finally, these practices are not limited to the nervous system and can be applied readily to any tissue and for studying numerous biological processes.

Acknowledgements

We thank our many lab members, past and present, who contributed to our development of genetic fate mapping techniques, especially GIFM, and for many lively discussions that provided the intellectual basis for our studies. We also thank Sandra Blaess, Charles Levine, and Roy Silitoe for their helpful comments on the manuscript. In addition, we thank other members of the scientific community who have contributed to refining the technical and conceptual advances in genetic fate mapping and GIFM. We have attempted to be comprehensive in our coverage of the subject and regret any unintentional omissions in references, ideas, or reagents.

Ancillary