All mouse lines were housed and fed according to the German Animal Welfare Act (Tierschutzgesetz) and the European Communities Council Directive in the Interfacultary Biomedical Facility (IBF), University of Heidelberg. The authorization for collecting mice brain and handling animal samples was approved by the Regierungspraesidium Karlsruhe, Baden-Wuerttemberg. The experiments were carried out in the Neuroanatomy Department, University of Heidelberg and all the procedures were performed according to the Animal Welfare Act. Adult mice were killed by cervical dislocation. The staining protocol for some antibodies required mice to be anesthetized with isofluorane and then intracardially perfused with fixative. For embryos, pregnant mice were sacrificed by cervical dislocation and E12.5, E15.5 or E18.5 pups were taken out of the uterus and processed for histology.

Phosphatase/NBT staining has been described (Lobe et al., 1999; Gierut et al., 2014). Briefly, brains or embryos were dissected in PBS buffer on ice; fixed in 4% paraformaldehyde solution with 0.02% NP-40 and 0.01% sodium deoxycholate at 4°C for 30 min; washed in PBS three times for 30 min at 4°C; incubated in PBS at 72°C for 30 min to inactivate endogenous alkaline phosphatase; rinsed three times in PBS for 10 min at room temperature; washed in alkaline phosphatase buffer two times 10 min; stained with 100 mg/ml NBT and 50 mg/ml BCIP in AP buffer at 4°C until optimal results appeared; washed in PBS extensively to reduce the background; mounted in Mowiol.

To detect specific cell markers as well as reporter proteins hPLAP and β-galactosidase, the following antibodies and conditions were used: anti-hPLAP (Sigma A2951) 1:2000 on paraffin sections (PS) or fixed frozen tissue (FFT); anti-β-galactosidase (Abcam ab9361) 1:200 (FFT); anti-Claudin11 (Abcam ab53041) 1:500 (FFT); anti-GalC (Chemicon AB142) 1:100 (PS); anti-myelin basic protein MBP (Sigma M3821) 1:100 (FFT); anti-platelet derived growth factor receptor alpha (PDGFRα; BD Pharmingen 558774) 1:300 on fresh frozen tissue; anti-NG2 (Chemicon AB5320) 1:100 (PS); anti-Olig2 (Chemicon AB9610) 1:200 (FFT); anti-GFAP (Chemicon AB5804) 1:300 (FFT); anti-β-tubulin III (Abcam ab18207) 1:600 (FFT); anti-NeuN (Abcam ab177487) 1:500 (FFT).

We cloned a DNA construct carrying a constitutive GFP reporter and a Cre-recombination-dependent tdTomato reporter. Every cell transfected will express the green GFP reporter, but transfected cells expressing Cre will express the red tdTomato reporter instead. Our construct was based on MSCV FLIPi P2G_Thy1.1 Dbl (p53, PTEN), a gift from Richard Hynes (Addgene plasmid #19746; Stern et al., 2008). We introduced the following modifications for our purposes: (1) removal of the cassette Puromycin 2a-GFP-Thy1.1-miR-WRPE; (2) insertion of the reporter tdTomato in reverse orientation in the place of Thy1.1-miR; (3) addition of the CAG promoter. The resulting DNA construct was transfected into the thalamus of E12.5 wild type mice through in utero electroporation. The electroporation technique has been described before (Tabata and Nakajima, 2001; Haddad-Tóvolli et al., 2013).

For proliferation studies, 50 mg/kg of body weight of Bromo-deoxy-uridine (BrdU) was intraperitoneally injected as aqueous solution into P10 mice three times, at 4 h intervals (Bu et al., 2004). Mice were sacrificed 2 h after the last injection. The brains were dissected and fixed in 4% PFA overnight at 4°C, washed in PBS for 4–6 h, cryoprotected in 0.1 M PBS containing 30% sucrose, embedded in OCT compound on Dry Ice and stored at −80°C. Previous to immunodetection of BrdU and Olig2, cryosections (20 μm) were incubated in 1 M HCl for 10 min on ice and then in 2 M HCl for 30 min at 37°C and neutralized with 0.1 M sodium borate buffer at room temperature.

The results were observed and analyzed under a confocal laser-scanning microscope (Zeiss LSM 510) with ZEN 2010 software.

The G-ratio was calculated on measurements performed on electron microscopy photographs of transverse sections of medulla oblongata (MO). The tissue samples were obtained from Foxb1^Cre/+ and Foxb1^Cre/Cre mouse brains (age P56, three mice per genotype), then processed according to current electron microscopy protocols. The diameter of 80 randomly chosen axons and the corresponding fibers (axon + myelin) were measured for each individual mouse. The data were analyzed with the Mann-Whitney test.

Cell countings were performed on 20 μm sections from the brains of three animals per genotype. Sagittal sections were cut in four series (A, B, C and D) from an area spanning the middle (rostro-caudally) of the hindbrain. Eight sections were counted per mouse. The counting bin for marker-labeled cells was 600 μm × 600 μm in MO. The counting bin for Olig2 plus BrdU was 0.1 mm² in MO. The localization of the bins was chosen randomly. Statistical assessment was performed with Prism 5 software (GraphPad Software, San Diego, CA, USA). The Mann-Whitney test was used and the results are represented in bar graphs with mean ± SD; *p < 0.05, **p < 0.01, ***p < 0.005.

Foxb1 is expressed in two narrow longitudinal bands in the VZ of the brainstem (Alvarez-Bolado et al., 1999) reminiscent of the domains of expression of OPC marker genes Olig1 and Olig2 (Zhou et al., 2000). We asked if the lineage of the Foxb1-expressing neuroepithelium included OL and if Foxb1 was involved in oligodendrocyte development. We approached this question by using a lineage-labeling mouse line, the Foxb1-Cre-EGFP x Z/AP. Foxb1-Cre-EGFP is a knockin-knockout of the Cre recombinase and the expression reporter EGFP into the Foxb1 locus (Zhao et al., 2007). By crossing these lines with the Z/AP reporter line (Lobe et al., 1999), Foxb1-expressing neuroepithelial cells as well as their entire progeny (even if they do not express Foxb1) will permanently express hPLAP in this way becoming identifiable. First, we examined the brains of mouse embryos heterozygous for this insertion. Foxb1-heterozygotes have normal phenotype (Dou et al., 1997; Alvarez-Bolado et al., 2000; Kloetzli et al., 2001) and were used here as proxies for the wild type. Treating early embryonic brains with a procedure to detect alkaline phosphatase activity, we detected as expected the bands in the hindbrain (Figures 1A–C). Later in development, we detected AP+ cells in the MO (Figure 1D).

By staining MO sections of these mice for hPLAP, we labeled large groups of cells with multiple processes in brains of every developmental age (Figure 2). As development proceeds, Foxb1-lineage cells form clusters in the brainstem as well as axon-like structures (Figures 2A–C). An internal control is built in the reporter mouse Z/AP (Lobe et al., 1999), so that cells not undergoing a Cre recombination (i.e., not expressing the Cre driver, in this case Foxb1) express β-galactosidase as a reporter instead. We labeled expression of both markers on brainstem sections to confirm that the hPLAP-expressing clusters did not overlap with β-galactosidase-expressing regions (Figures 2D–F).

In order to identify the cellular components of these clusters, we colocalized hPLAP with neuron-specific markers (either β-tubulin-III or NeuN) on Foxb1^Cre/+ x Z/AP brains and found a number of double-labeled cells (Figures 3A–F). We confirmed this on primary culture of Foxb1^Cre/+ x Z/AP brain cells (Figures 3G–I). Colocalizing astrocyte-specific protein glial fibrillary acidic protein (GFAP) with hPLAP also showed a small number of Foxb1-lineage astrocytes (not shown). As expected, no hPLAP-labeled cells expressed microglial marker Iba-1, since microglia originates outside of the neural tube and therefore far from the Foxb1-expressing neuroepithelium (not shown).

The finding that only small numbers of hPLAP-labeled brain cells were neurons or astrocytes suggested that the most abundant cellular component of the Foxb1 lineage could be OL. Expression of myelin basic protein (MBP) colocalized partially with hPLAP (Figures 4A–C), in principle confirming this impression. The appearance of the tissue was however complex and did not provide appropriate cellular resolution (Figures 4A–C). Expression of Claudin11, a marker of mature OL (Bronstein et al., 2000), colocalized with profiles morphologically similar to myelinated axons (Figures 4D–F). On primary cultures of cells from the MO of Foxb1^Cre/+ x Z/AP mice, Claudin11 and hPLAP colocalized on cells with unequivocal mature oligodendrocyte morphology, including the characteristic lamellae (Figures 4G–I). Expression of Galactocerebroside C (GalC), a marker of mature as well as immature OL, also colocalized with hPLAP on sections (Figures 4J–L) and on primary cultures (Figures 4M–O).

If most of the observed Foxb1-lineage brain cells are OL, it follows that some OPC must also belong to this lineage. To confirm this, we colocalized Foxb1-lineage reporters with OPC markers Olig2 and NG2 (Figure 5). Since Olig2 protein localizes to the cell nucleus, in order to colocalize it with a reporter we crossed our Foxb1^Cre/+ x Z/AP mice with the reporter mouse line ROSA26R (Soriano, 1999), whose cells express β-galactosidase as reporter upon Cre recombination. We found numerous cells expressing both this reporter and Olig2 (Figures 5A–C) in the MO of Foxb1^Cre/+ x ROSA26R mice. The β subunit of the receptor for PDGFRβ as well as NG2 specifically colocalize in OPC (Richardson et al., 1988; Hart et al., 1989). We found also many cells where NG2 colocalized with hPLAP in histological sections of the MO (Figures 5G–I) as well as in culture (Figures 5J–L; see cell abundance quantitation below, Figure 8).

OPC originate from certain spatially restricted regions of the neural tube. They then migrate and proliferate to supply the entire CNS with OL. A current question concerns the possibility that OL from different origins have differential properties and if OL from any given region can be used to substitute for OL in any other region of the brain (replacement therapy). In this context, we asked if the basic composition of the Foxb1 lineage in the hindbrain was similar to that of another regionally restricted subpopulation of Foxb1-lineage brain cells situated not in the hindbrain but in the forebrain. The thalamus is a forebrain region with strong Foxb1 developmental expression and a large part of the thalamus is of Foxb1 lineage (Zhao et al., 2007, 2008).

Clusters expressing hPLAP were obvious in the thalamus (Figure 6A). A characteristic property of the thalamus with important functional consequences is the presence of many neurons sending axonal projections to the cortex. We used in utero electroporation with a plasmid expressing upon Cre recombination the red reporter tdTomato (see “Materials and Methods” Section for details), which fills the cell body and axons. The results showed abundant Foxb1-lineage thalamo-cortical neurons (Figure 6B). We confirmed the presence of OL, neurons and astrocytes colocalizing hPLAP on primary thalamic culture (Figures 6C–K; see cell abundance quantitation below, Figure 9).

Comparison of Foxb1^Cre/+ x Z/AP and Foxb1^Cre/Cre x Z/AP mouse brains after whole mount histochemical demonstration of AP activity showed very increased staining in the homozygotes (Figures 7A–D). This suggested that Foxb1 had an influence on the cell abundance and maybe also identity. To investigate this, we performed cell countings on tissue of heterozygotes and homozygotes labeled with specific cell markers as well as hPLAP. Our results show, first, that in homozygotes there are more Foxb1-lineage cells (Figure 8A). We also found that, in homozygotes, Foxb1-lineage neurons, astrocytes and mature OL were reduced (Figures 8B–E). However, OPC were increased (Figures 8F,G). Interestingly, a first approach to the abundance of different Foxb1-lineage cell types in the thalamus yielded similar results (Figures 9A–D). This suggests that all Foxb1-lineage OL are functionally homogeneous.

Although the exact proportion of Foxb1-lineage cells to other brain cells is not central to our study, we thought it was of some interest to obtain at least a rough approximation to this value for neurons, OL and OPC. Therefore we counted non-hPLAP (i.e., non-Foxb1-lineage) cell in the bins used to quantitate the hPLAP-expressing (Foxb1-lineage) cells. The results are as follows: (1) mature plus immature OL (i.e., GalC-expressing cells) belonging to the Foxb1-lineage are 73.17% of the total GalC-positive cells per bin in heterozygotes, and 90.00% in homozygotes; (2) OPC (i.e., NG2-expressing cells) belonging to the Foxb1-lineage are 45.40% of all NG2-positive cells per bin in heterozygotes and 75.89% in homozygotes; and (3) neurons (i.e., cells expressing beta-tubulin III) belonging to the Foxb1-lineage are 30% of all beta-tubulin III-expressing cells per bin in heterozygotes and 24.78% in homozygotes. As could be expected, OL and OPC of the Foxb1 lineage “crowd out” the non-Foxb1 lineage ones in homozygotes, while the proportion of Foxb1-lineage neurons becomes slightly smaller. These data however cannot be directly compared to our cell countings summarized in Figure 10, since those are represented in percent of hPLAP-expressing cells (as opposed to percent of all cells in the bin).

We concluded that, in the Foxb1-deficient MO, the OPC increase to the expense of neurons and mature OL (Figure 10).

In order to confirm this increase in OPC, we wanted to use another OPC marker, Olig2. Since Olig2 is localized in the nucleus, we wanted a reporter of the Foxb1 lineage also localized in the nucleus. For this reason we used a different reporter, the ROSA26R mouse line (Soriano, 1999). In this line, Cre recombination causes expression of β-galactosidase which can be seen in the nucleus. We colocalized Olig2 and β-galactosidase on MO sections of heterozygotes and homozygotes (Figures 11A–F). Cell counting confirmed a significant increase of Olig2+ cells in the homozygotes (Figure 11G). Additionally, we labeled dividing cells with BrdU (see “Materials and Methods” Section for details) and then labeled the tissues with BrdU and Olig2 (Figure 12). We found that, in homozygous mutants, more Olig2-expressing cells were dividing than in heterozygotes.

Finally, we asked if there would be any obvious defect in myelination in the hindbrain of our mutants. The phenotype of Foxb1-deficient mice includes several behavioral and CNS developmental defects (Dou et al., 1997; Labosky et al., 1997; Alvarez-Bolado et al., 2000; Radyushkin et al., 2005; see “Discussion” Section). However, alterations suggesting myelination defects have not been described in these mice. This suggests that the OL in this mutant are able to myelinate correctly. To test this hypothesis, we measured axon and fiber diameter on electron microscopy images from the hindbrain of homo- and heterozygotes and found that the G-ratio (Friede and Bischhausen, 1982) does not change in the mutants (Figure 13).