Decreased ACKR3 (CXCR7) function causes oculomotor synkinesis in mice and humans

Abstract

Oculomotor synkinesis is the involuntary movement of the eyes or eyelids with a voluntary attempt at a different movement. The chemokine receptor CXCR4 and its ligand CXCL12 regulate oculomotor nerve development; mice with loss of either molecule have oculomotor synkinesis. In a consanguineous family with congenital ptosis and elevation of the ptotic eyelid with ipsilateral abduction, we identified a co-segregating homozygous missense variant (c.772G>A) in ACKR3, which encodes an atypical chemokine receptor that binds CXCL12 and functions as a scavenger receptor, regulating levels of CXCL12 available for CXCR4 signaling. The mutant protein (p.V258M) is expressed and traffics to the cell surface but has a lower binding affinity for CXCL12. Mice with loss of Ackr3 have variable phenotypes that include misrouting of the oculomotor and abducens nerves. All embryos show oculomotor nerve misrouting, ranging from complete misprojection in the midbrain, to aberrant peripheral branching, to a thin nerve, which aberrantly innervates the lateral rectus (as seen in Duane syndrome). The abducens nerve phenotype ranges from complete absence, to aberrant projections within the orbit, to a normal trajectory. Loss of ACKR3 in the midbrain leads to downregulation of CXCR4 protein, consistent with reports that excess CXCL12 causes ligand-induced degradation of CXCR4. Correspondingly, excess CXCL12 applied to ex vivo oculomotor slices causes axon misrouting, similar to inhibition of CXCR4. Thus, ACKR3, through its regulation of CXCL12 levels, is an important regulator of axon guidance in the oculomotor system; complete loss causes oculomotor synkinesis in mice, while reduced function causes oculomotor synkinesis in humans.

Introduction

Binocular vision and social communication require precise control of eye movements by three cranial nerves (CNs) innervating six extraocular muscles (EOMs). The oculomotor nerve (CN3) innervates the medial, inferior and superior rectus muscles and the inferior oblique and levator palpebrae superioris muscles, the trochlear nerve (CN4) innervates the superior oblique muscle and the abducens nerve (CN6) innervates the lateral rectus muscle. Failure to innervate EOMs can result in limited eye movements and/or ptosis (drooping eyelid), while miswiring between these nerves and muscles can lead to oculomotor synkinesis, the involuntary movement of the eyes or eyelid with a voluntary attempt at another eye or facial movement. When developmental, these disorders are referred to as the congenital cranial dysinnervation disorders (CCDDs) (1).

The most common forms of oculomotor synkinesis are Marcus Gunn jaw winking (MGJW) and Duane retraction syndrome (DRS). MGJW, in which a ptotic eyelid will elevate with chewing or sucking, is believed to be caused by aberrant innervation of the levator palpebrae superioris (which elevates the eyelid) by nerve fibers of the motor branch of the trigeminal nerve, which normally innervate the muscles of mastication. DRS, in which abduction of the eye is limited and globe retraction occurs with adduction, results from the absence of CN6 and aberrant innervation of the lateral rectus by axons from CN3 (2, 3). MGJW, globe retraction and other forms of oculomotor synkinesis can occur in isolation or in conjunction with other CCDDs (1).

We have recently shown that Marcus Gunn-like aberrant innervation of the EOMs occurs in mice with loss of the chemokine receptor CXCR4 or its ligand CXCL12 (4). Loss of CXCR4/CXCL12 signaling leads to the failure of CN3 axons to exit the midbrain neuroepithelium ventrally; axons instead grow dorsally and stall. As a result, CN3 target EOMs do not receive their normal innervation and are instead aberrantly innervated by misdirected axons of the motor trigeminal nerve.

CXCR4/CXCL12 signaling is regulated by a related atypical chemokine receptor, ACKR3 (also known as CXCR7), which also binds CXCL12 (5) and can act as a scavenger receptor (6–8). Unlike CXCR4, ACKR3 does not signal through G-proteins, although it can recruit beta-arrestin (9). Instead, binding of CXCL12 to ACKR3 leads to ligand internalization and degradation, thereby reducing extracellular levels of CXCL12 (5). ACKR3 is widely expressed in mouse brain, including in neural progenitors (10).

Ackr3 knock-out mice are reported to have significant cardiovascular defects with high perinatal lethality (11, 12). Those that survive have cardiac defects but reach adulthood and are fertile. While the developing nervous system was initially reported as normal (12), later studies revealed multiple roles of ACKR3 in brain development. In neural progenitor cells, ACKR3 functions in both proliferation (13) and migration (14). In migrating cortical interneurons, loss of Ackr3 leads to decreased expression of CXCR4 protein (15). This is because ACKR3 sequesters CXCL12 and, in the absence of ACKR3, there is excess CXCL12. Excessive CXCR4–CXCL12 binding subsequently leads to CXCR4 downregulation (8), because CXCR4 protein is phosphorylated and degraded in response to CXCL12 (16). Thus, proper migration of cortical interneurons requires CXCL12 and both CXCR4 and ACKR3 receptors. CXCR4 and ACKR3 are also both required for maintenance of the radial glial scaffold and PNS/CNS boundary in the spinal cord (17), proper migration of facial motor neurons in the hindbrain (18) and guidance of trigeminal sensory axons (19).

We have identified a homozygous missense variant in ACKR3 in a consanguineous family that segregates ptosis and a unique form of oculomotor synkinesis—elevation of a ptotic eyelid with abduction of the ipsilateral eye—as a recessive trait. We show that the missense variant decreases the binding affinity of ACKR3 for CXCL12. Mice lacking Ackr3 show axon guidance abnormalities of CN3, CN4 and CN6. Given the role of CXCR4 in CN3 development (4), we asked whether ACKR3 also regulated CXCR4 in the ocular motor system. We found that loss of Ackr3 leads to degradation of CXCR4 protein and that excess CXCL12 (as presumably occurs in the absence of ACKR3) leads to oculomotor axon misrouting, similar to loss of CXCR4 signaling.

Results

Family with ptosis and oculomotor synkinesis segregated a homozygous missense variant in ACKR3.

We ascertained a consanguineous family from Saudi Arabia that segregated variable ptosis and synkinetic lid elevation on ipsilateral abduction as a recessive trait in three siblings and a double cousin (Fig. 1A–E); the branch with three affected siblings was previously reported (20). On examination of the three siblings, IV-3 had bilateral congenital ptosis without apparent synkinesis, IV-9 had bilateral ptosis with elevation of the eyelid on ipsilateral abduction (Fig. 1A–D) and IV-13 had left congenital ptosis, ipsilateral lid elevation on abduction and slight limitation of adduction and globe retraction with adduction. The previously unreported cousin, IV-14, exhibited right congenital ptosis with ipsilateral lid elevation on abduction. The ptosis and synkinesis in these four individuals was most consistent with failure of innervation of the levator palpebrae superioris by CN3 and its aberrant innervation by CN6. Interestingly, a half-sister of the three affected siblings, IV-1, had right-sided Duane syndrome with limitation of abduction, mild limitation of adduction and globe retraction with upshoot of the right eye on attempted adduction. Mouse models of Duane syndrome have revealed that this pattern of synkinesis results from absence of CN6 and aberrant innervation of the lateral rectus by axons of CN3 (2, 3).

Figure 1

Family with oculomotor synkinesis segregates a homozygous ACKR3 c.772G>A (p.Val258Met) variant. (A–D) Photographs of affected participant (IV-9) showing bilateral ptosis in primary gaze (A) with lid elevation with ipsilateral abduction (B,C) and minimal lid elevation on upgaze (D). (A–D) reprinted with permission from Khan et al., JAAPOS 2004. (E) Family pedigree. Black-filled circles indicate affected individuals: IV-3 has bilateral congenital ptosis, IV-9 has bilateral congenital ptosis with lid elevation on abduction, IV-13 has left sided ptosis with ipsilateral lid elevation on abduction and IV-14 has right-sided ptosis with ipsilateral lid elevation on abduction. All are homozygous for the ACKR3 c.772G>A (p.Val258Met) variant. The half-sister filled in gray (IV-1) has isolated Duane syndrome without ptosis. She is heterozygous for the variant. The parents (III-2 and III-3) are heterozygous for the variant and are unaffected by any oculomotor abnormality. Genotypes for the ACKR3 variant and chromosome 2 polymorphic markers are shown underneath each individual. Two-point linkage analysis at D2S338 reveals a LOD score of 3.3. (F) Sanger sequencing of ACKR3 shows c.772G in control, heterozygous c. 722G>A in the unaffected mother and homozygous c.722G>A in an affected individual. (G) Valine258 is conserved in mammals and X. tropicalis, but not D. rerio, in which the gene is duplicated. (H) Schematic of ACKR3. Val258 is in the sixth transmembrane domain (red). Extracellular cysteines are labeled in yellow. K208, D179 and D275 (purple) are necessary for CXCL12 binding (33) (see Discussion). Residues labeled in pink were mutated in (33) but did not decrease CXCL12 binding. Residues labeled in green have homozygous missense variants present in the gnomad database.

Homozygosity analysis using SNPs from the four affected individuals (IV-3, IV-9, IV-13, IV-14) and two unaffected parents (III-02, III-03) revealed only one 4.4 Mb region on chromosome 2 (hg19:Chr2: 234937171-239369790) that was homozygous in the affected individuals and heterozygous in the unaffected parents. Multipoint linkage analysis identified two regions consistent with linkage: a 3 Mb region on chromosome 2 (Chr2:235547435-238639192) that fell within the region of homozygosity and a 6 Mb region on chromosome 7ter (Chr7:152081816-158383513), with maximum logarithm of the odds (LOD) scores of 2.38 and 2.36, respectively. Two-point linkage analysis including the seven unaffected siblings (Fig. 1E) of markers encompassing the chr 2 region revealed a maximum LOD score of 3.3, theta 0, at marker D2S338 (chr2:237135394-237335743, hg19) (Fig. 1E).

Sequence analysis of coding and intron-exon boundaries of all genes that fell within the chromosome 2 linkage region using both Sanger and next generation sequencing (whole-exome sequencing of IV-9 and IV-14 and whole-genome sequencing of IV-13) identified only one rare coding variant which segregated with affection status: a homozygous c.772G>A (NM_020311.2) missense variant in ACKR3 (CXCR7) (Fig. 1F). Notably, the half-sister who had Duane syndrome without ptosis (IV-1) was heterozygous for this variant (Fig. 1E, gray circle in pedigree). Analysis of all rare non-coding variants that fell within the chromosome 2 linkage region revealed 70 single nucleotide variants (SNVs), none of which altered residues conserved between species or residues located within known regulatory regions. The region of linkage contained no deletions or duplications.

To rule out compound heterozygous variants in the chromosome 7ter region that were consistent with linkage but not homozygous in affected individuals, we analyzed additional markers within and centromeric to this region in the seven unaffected siblings. The maternal and paternal haplotypes present in the affected individuals were also present in one of the unaffected siblings, ruling out Chr7ter linkage. Consistent with this, there were no rare coding variants in the region that segregated with affection status. Finally, analysis of the two exomes and the genome did not reveal any other shared rare SNVs or duplications across the genome.

The ACKR3 c.772G>A variant is predicted to be disease causing by Mutation Taster (21), damaging by Sift (22), probably damaging by Polyphen (23), tolerated by Fathmm (24) and has a CADD Phred score (25) of 28.3 (damaging). The variant is not present in either a heterozygous or homozygous state in the 1000 genomes, ExAC (26) or gnomAD databases. The variant is predicted to result in a p.Val258Met amino acid substitution in the sixth transmembrane domain of ACKR3 (Fig. 1H). The Val258 residue is conserved in mammals and X tropicalis, but not Danio rerio, in which the gene is duplicated (Fig. 1G).

To identify additional disease alleles, 581 unsolved pedigrees with isolated congenital ptosis, MGJW or other CCDDs were screened for variants in ACKR3, CXCR4 and CXCL12, and no rare variants were found. None of these probands, however, had ipsilateral lid elevation with abduction.

The V258M mutant protein was expressed and trafficked to the cell surface in vitro

To determine if the mutant V258M-ACKR3 protein had altered expression, folding or intracellular localization compared to wild-type ACKR3 protein, plasmids encoding human wild-type or c.772G>A (p.V258M) ACKR3 were transiently transfected into HEK293 cells. Staining with anti-human ACKR3 antibody revealed equal expression of wild-type and mutant receptors and both trafficked to the cell surface (Fig. 2A–C).

Figure 2

V258M ACKR3 is expressed and traffics to the cell surface, but it has a lower Kd for CXCL12 binding. HEK293 cells were transiently transfected with plasmids encoding wild-type human ACKR3 (A), V258M human ACKR3 (B) or empty vector (C), fixed after 24 h and stained with anti-human ACKR3 antibody (red). Expression and localization of the mutant protein is indistinguishable from wild-type. Binding affinity for CXCL12 was tested using an HTRF assay (see Materials and Methods for explanation). Kd for the wild-type receptor is 0.2028 nM ±0.06792 (mean ± SEM), similar to published reports of the Kd for ACKR3 and CXCL12 binding (7, 27, 28). Kd for the mutant receptor is 3.814 ± 1.214 (P = 0.0267, paired t-test, n = 6).

The V258M mutant protein had lower binding affinity for CXCL12 in vitro (higher Kd)

Binding affinity of the wild-type and mutant ACKR3 receptors for the CXCL12 ligand was evaluated using an Homogeneous Time Resolved Fluorescence (HTRF) assay. In this assay, the receptor is expressed on the cell surface with a SNAP-tag, which is then labeled with a fluorescent protein that fluoresces at 620 nm. Ligand labeled with a second fluorescent tag is added, and when ligand and receptor bind, the two fluorescent tags interact and fluoresce at 655 nm. Using this assay, we observed that both the wild-type and mutant receptors could bind CXCL12, suggesting that the mutation did not cause complete loss-of-function. However, the mutant receptor bound with a lower affinity. We calculated a Kd of 0.2028 nM ±0.06792 (mean ± SEM) for the wild-type receptor, similar to published reports of the Kd for ACKR3 and CXCL12 binding (7, 27, 28) (Fig. 2D). By contrast, the mutant receptor had a Kd of 3.814 ± 1.214, (P = 0.0267, paired t-test, n = 6) indicating the mutant receptor binds CXCL12 with significantly lower affinity than does the wild type receptor.

Ackr3 was widely expressed in embryonic mouse

We next examined expression of Ackr3 during mouse development. In situ hybridization revealed diffuse expression of Ackr3 mRNA throughout the embryo (not shown), including in the developing neuroepithelium in the region of the developing oculomotor nucleus at E11.5 (Fig. 3A, B) and E13.5 (Fig. 3C, D). Ackr3 embryonic mouse knock-out tissue (Ackr3^KO/KO:Isl^MN-GFP) showed no Ackr3 expression (Fig. 3E, F).

Figure 3

Expression of Ackr3 mRNA in region of developing mouse oculomotor nucleus. In situ hybridization of Isl1 on wild-type (A, C) and Ackr3^KO/KO (E) tissue shows Isl1 expression in developing oculomotor neurons at E11.5 (A, E) and E13.5 (C). In situ hybridization of Ackr3 on wild-type (B, D) and Ackr3^KO/KO (F) tissue shows Ackr3 expression in wild-type Isl1-expressing CN3 neurons and surrounding tissues at E11.5 (B) and E13.5 (F) and absence of expression in Ackr3^KO/KO tissue (H). All Isl1 and Ackr3 pairs are adjacent sections. Scale bar in (G) equals 100 µm for all sections.

Figure 4

Ackr3 knock-out mice display variable misrouting of CN3, CN4 and CN5. (A) E11.5 maximum intensity projections of whole mount imaging of an Isl^MN:GFP wild-type embryo, counterstained with anti-smooth muscle actin to label muscles and arteries (red), demonstrates the normal trajectory of the oculomotor (CN3), trochlear (CN4), abducens (CN6), motor trigeminal (CN5m) and facial (CN7) nerves. Midbrain to orbit of one side of embryo enlarged in (B). (D) E11.5 Ackr3^KO/KO embryo. Oculomotor axons project dorsally in the midbrain and stall. CN7, CN9, CN10 and CN12 have grossly normal trajectories, while spinal cord axons project dorsally and aberrantly. There is variability in CN3, CN4, CN5m and CN6 trajectories, as displayed in (D), (E) and (G–I). In (D) and (E), all oculomotor axons project dorsally and stall, while in (G) and (H), axons from the caudal portion of the oculomotor nucleus project dorsally, but the rostral portion forms a thin CN3, which projects to the orbit. The arrow in (H) shows an abnormal branch of CN3. In 8/24 orbits, axons of CN5m aberrantly project toward the orbit [arrow in (G)]. CN6 reaches the orbit in 8/24 orbits (E), stalls in 4 (G) and cannot be identified in 12/24 orbits (D, H). (C) Photomicrograph of a whole E11.5 wild-type embryo. (F) Two embryos displayed severe malformations of the back of the head. In these embryos, CN4 is severely misrouted, reaching the midline, but not crossing and instead projecting caudally (I). Images in (B), (E) and (G–I) have been cropped in the z dimension to show only one side of the head. N = 12 Ackr3^KO/KO and 10 wild-type controls. Scale bars: 500 µm (A, D, I) and 200 µm (B, E, G, H).

Figure 5

Loss of Ackr3 leads to aberrant midline crossing of CN3 axons rostrally. (A) Dorsal view of CN3 and CN4 in an E11.5 wild-type Isl^MN-GFP embryo, counterstained with anti-smooth muscle actin to label muscles and arteries (red), reconstructed and cropped using Imaris software from a whole mount embryo imaged sagitally. CN3 projects ventrally (into the page, with no midline crossing). CN4 projects dorsolaterally and crosses the midline before projecting contralaterally. (B) Dorsal view of an E11.5 Ackr3^KO/KO embryo shows rostral midline crossing of CN3 axons (arrow). CN4 projects normally. Scale bar: 400 µm.

Loss of ACKR3 caused variable oculomotor and motor trigeminal axon misrouting and could cause DRS pathology

Examination of cranial motor neurons and axons in twelve E11.5 whole mount Ackr3^KO/KO:Isl^MN-GFP embryos showed significant variability in axon trajectories (Fig. 4). All showed misrouting of the oculomotor nerve, but to varying degrees. In 5/24 CN3 nuclei, all CN3 axons projected dorsally (and stalled), rather than ventrally (Fig. 4D, E, Video 1). Some of the dorsally projecting axons also grew rostrally and appeared to cross the midline earlier and more rostrally than normal CN3 cell body midline crossing (Fig. 5, Video 2). In the remaining 19/24 CN3 nuclei, axons from the caudal portion of the nucleus projected dorsally in the brainstem and stalled, while axons from the rostral portion exited the brainstem ventrally and formed a thin nerve (mean width 6.7 ± 0.6 µm for KOs with a nerve versus 14.6 ± 0.4 µm for wild-type nerves, P < 0.0001, unpaired t-test), which projected toward the orbit (Fig. 4G, H). Seven of these nineteen CN3 nerves had abnormal trajectories, five had abnormal branching (Fig. 4H) and one appeared to stall on the way to the orbit. On 8/24 sides, aberrant branches from the motor trigeminal nerve could be visualized extending to the orbit, suggestive of a MGJW-like pathology (Fig. 4C). CN4 had a normal trajectory in most E11.5 mutant embryos. Two of the 12 embryos, however, had a cranial malformation with apparent collapse of the fourth ventricle (Fig. 4C, F), and in these 2 embryos CN4 projected dorsally but did not cross the midline and instead turned caudally (Fig. 4I). Heterozygous animals displayed normal nerve trajectories.

The CN3 phenotype of Ackr3^KO/KO:Isl^MN:GFP embryos at E13.5 and E14.5 continued to be variable. CN3 axons projected dorsally, stalled in the midbrain and could not be identified in the periphery of 4/8 embryos bilaterally and 3/8 embryos unilaterally. In the eighth embryo, CN3 was present bilaterally but very thin, and axons of CN3 and CN6 appeared to touch near the orbital apex (not shown). In the three embryos in which CN3 was present in the periphery unilaterally, CN3 was thin and CN6 was absent on the same side. In these three embryos, the thin CN3 projected to its normal EOM targets but also aberrantly projected to the lateral rectus (which lacked innervation by the absent CN6), as seen in Duane syndrome (Fig. 6A, B). Notably, all the aberrant projections to the lateral rectus came from the inferior decision region (29) of CN3.

Figure 6

Aberrant innervation patterns of CN3 and CN6. (A) View of orbit in an E13.5 Isl^MN:GFP:Hb9:GFP embryo, counterstained with anti-smooth muscle actin to label muscles and arteries (red), reconstructed and cropped using Imaris software from a whole mount embryo imaged sagitally, shows the normal innervation patterns of CN3, CN4 and CN6. CN3 enters the orbit, the superior division branches to the SR (yellow arrowhead), an inferior decision region forms between the MR and IR (yellow arrow), and a branch extends to the IO. CN4 projects to the SO. CN6 projects to the LR. (B) Example of an Ackr3^KO/KO with Duane syndrome pathology. CN6 is absent. CN3 is thin, has scant projections to the SR and aberrant branches to the LR (yellow arrow). (C) Central orbital view of an E13.5 Hb9:GFP wild-type embryo shows the normal trajectory of CN6 to the LR. (D) Example of an E13.5 Ackr3^KO/KO embryo in which CN6 not only innervates the LR muscle but also has an aberrant branch that divides to innervate other EOMs that are normally innervated solely by CN3 (yellow arrow). Scale bar in (A) equals 200 µm in (A, B). Scale bar in (C) equals 200 µm in (C, D). SR, superior rectus; SO, superior oblique; MR, medial rectus; IR, inferior rectus; IO, inferior oblique; LR, lateral rectus.

Ackr3^KO/KO abducens nerves ranged from absent to aberrant to normal

To better assess the trajectory of CN6 in Ackr3^KO/KO embryos, we crossed them to Hb9:GFP reporter mice, which have strong expression of GFP in CN6. Of the 6 E13.5 Ackr3^KO/KO:Hb9:GFP embryos examined, CN6 was absent in 5 of 12 and very thin in 1 of 12 orbits. CN6 was visible in the remaining six orbits; it had a normal trajectory in two and displayed aberrant branching within the orbit in four (Fig. 6C, D).

Ackr3^KO/KO facial nerves had normal peripheral branching but abnormal cell body migration

In wild-type mice, facial motor neurons are born in rhombomere 4 and their axons exit the hindbrain at that level, turn, exit the skull at the stylomastoid foramen and then divide into five main branches. As the axons are extending, the cell bodies migrate within the hindbrain from rhombomere 4 to rhombomere 6. This leads to the formation of the internal genu, a sharp turn of the axons where the cell bodies were originally born (Fig. 7A). A second population of motor neurons, the inner ear efferents (IEEs), are born in the same progenitor zone as the facial motor neurons in rhombomere 4, and then migrate laterally and extend their axons to the inner ear (30). In E14.5 Ackr3^KO/KO embryos, the facial nerve followed a normal trajectory, and all five major divisions were present and branched normally (Fig. 7C and D). The IEE cell bodies also had a normal lateral trajectory. The caudal migration of the facial motor neuron cell bodies, however, was perturbed. Similar to other Ackr3^KO/KO phenotypes, there was variability between embryos, ranging from no cell body migration (Fig. 7B) to embryos with apparently normal migration unilaterally. These data are consistent with that reported in zebrafish, in which knock-down of cxcr7b disrupts facial motor neuron migration from rhombomere 4 to rhombomere 6, leading to an accumulation of cell bodies in rhombomere 5 (18).

Figure 7

The facial nerve branches normally, but cell bodies in the facial nucleus fail to migrate. (A) Sagittal view of hindbrain in an E13.5 Isl^MN:GFP embryo reconstructed and cropped using Imaris software from a whole mount embryo imaged sagitally shows the trajectory of the facial nerve, inner ear efferents and the migrating cell bodies (yellow arrow). (B) In an E13.5 Ackr3^KO/KO embryo, the facial nerve and IEEs have normal projections from the hindbrain, but the cell bodies do not migrate (yellow arrow). (C) Sagittal view of an E14.5 Isl^MN:GFP embryo shows the normal branching pattern of CN7. (D) In an E13.5 Ackr3^KO/KO embryo, the facial nerve shows normal peripheral branching. Scale bar in (A) equals 200 µm for (A, B) and scale bar in (C) equals 300 µm for (C, D).

Ackr3^KO/KO embryos downregulated CXCR4 protein, but not Cxcr4 mRNA

In cortical interneurons, loss of ACKR3 was reported to cause excess extracellular levels of CXCL12 and resulting ligand-dependent downregulation of CXCR4 protein (8, 15, 16). To determine if a similar mechanism was at work in oculomotor neurons, we examined expression of CXCR4 protein and Cxcr4 mRNA in Ackr3^KO/KO tissue. While levels of Cxcr4 mRNA were similar between wild-type and knock-out oculomotor nuclei (Fig. 8A, B, F, G, K), there were lower levels of CXCR4 protein expression in the oculomotor nucleus and surrounding tissue (Fig. 8C–E, H–J) and whole brain (Fig. 8L and M) in knock-out compared to wild-type tissues (Fig. 8), indicating regulation at the level of protein, rather than transcription.

Figure 8

CXCR4 protein, but not mRNA, is downregulated with loss of Ackr3. In situ hybridization for Isl1 (A) and Cxcr4 (B) and immunohistochemistry for Isl^MN:GFP (C, E; green) and CXCR4 (D, E; red) at E11.5 shows Cxcr4 expression in and around CN3. In Ackr3^KO/KO embryos, in situ hybridization for Isl1 (F) and Cxcr4 (G), and RT-PCR for Cxcr4 (K) shows normal levels of Cxcr4 mRNA, but immunohistochemistry for Isl^MN:GFP (H, J; green) and CXCR4 (I, J; red) and western blot of whole brain for CXCR4 (L, M) show very low levels of CXCR4 protein expression. (A) and (B) are adjacent sections, as are (F) and (G). Scale bar in (A) equals 100 µm for (A, B, F, G). Scale bar in (H) equals 100 µm for (C–E, H–J). In (K), differences are not significant (P = 0.69). In (M), asterisks indicate P < 0.001.

Excess CXCL12 caused a similar phenotype to inhibition of CXCR4

To test the hypothesis that the ocular motor phenotypes seen with loss of Ackr3 result from excess CXCL12 levels (due to loss of ACKR3’s scavenging function) and secondary downregulation of CXCR4 protein, we tested the effect of the addition of excess CXCL12 on CN3 development in an ex vivo embryonic slice assay (4, 31). Wild-type E10.5 Isl^MN:GFP embryos were harvested, embedded in agarose and sliced so that the oculomotor nucleus and eye were in the same slice. Slices were grown on tissue culture inserts in a stage-top incubator and time-lapse photography was taken every 30 min. When excess CXCL12 was added to the growth media at concentrations of 50 or 100 nM, oculomotor axons that had not yet left the nucleus projected dorsally, rather than ventrally (Fig. 9). Axons that had already exited the midbrain continued toward the eye. This phenotype was strikingly similar to the phenotype seen with inhibition of CXCR4 by the small molecule inhibitor AMD3100 (4) and similar to that found in vivo in Cxcr4 and Ackr3 knock-out animals.

Figure 9

Excess CXCL12 causes CN3 axon misrouting, similar to inhibition of CXCR4. Time-lapse imaging of slice cultures from E10.5 Isl^MN:GFP embryos at 0, 12, 24 and 36 h in culture. In wild-type embryos, (top) over the course of 36 h CN3 (green) grows toward the eyes and branches extensively. When 50 or 100 nM CXCL12 is added (middle panels), axons exit the oculomotor nucleus dorsally. Those that had already exited continue toward the eye. This phenotype is strikingly similar to the phenotype when 1 µg/ml AMD3100 (specific inhibitor for CXCR4, bottom) is added. Bottom panel reprinted with permission from Whitman et al. IOVS 2018. D, dorsal; V, ventral; E, eye; N, oculomotor nucleus; arrow, oculomotor axons (white: wild-type projection, yellow: aberrant projection). Scale bar equals 200 µm.

Discussion

We have identified a consanguineous pedigree in which affected members have congenital ptosis accompanied by an unusual form of oculomotor synkinesis and who segregate a homozygous ACKR3 missense variant. While we have identified a rare ACKR3 variant only in a single pedigree, our genetic, functional and developmental data support its pathogenicity.

The disorder in the pedigree is linked to a 3 Mb region of chromosome 2 with a maximum LOD score of 3.3; this is also the only region of significant homozygosity. Within this region, ACKR3 c.772G>A is the only rare coding variant identified by candidate gene and whole-exome sequencing. Moreover, by analysis of whole-genome sequence, the region of linkage contains no structural variants and no non-coding variants in highly conserved regions. Finally, there are no coding duplications elsewhere in the genome that co-segregate with the phenotype, supporting the lack of an insertion into this region. The ACKR3 variant is absent from the gnomAD database (26), which contains homozygous missense variants at only seven residues (mapped in Fig. 1H) and no individuals with homozygous loss of function variants. The ACKR3 gene is constrained against loss-of-function variants, with an o/e (observed/expected ratio) of 0.22 (90% confidence interval, 0.09–0.70) and a pLI (probability of being loss-of-function intolerant) (26) of 0.38. Complete loss of Ackr3 function in mice causes severe systemic phenotypes, including cardiac defects and perinatal death (11, 12), likely explaining why ACKR3 is constrained against loss-of-function variants in humans. ACKR3 is mildly constrained against missense variants, with a Z-score of 1.01 and o/e ratio of 0.81 (0.72–0.92) and is not constrained against synonymous variants, with a Z-score (32) of −0.42 and o/e ratio of 1.05 (0.90–1.24). The small size of the gene and thus low number of expected variants, however, makes the 90% confidence intervals large and constraint difficult to interpret. Given the loss of function constraint and the severe phenotypes present in mice with loss of Ackr3, we hypothesize that ACKR3 c.772G>A is a hypomorphic allele that causes a very precise, mild decrease in ACKR3 function to which development of oculomotor and abducens axons are selectively vulnerable.

The ACKR3 c.772G>A variant, predicted to cause a V258M amino acid substitution, leads to a functional protein with a lower binding affinity for the ligand CXCL12 in vitro, consistent with it acting as a hypomorphic allele. Benredjem et al (33) assessed the effects of selected ACKR3 mutations on binding to CXCL12 and CXCL11. Although mutations altering V258, which falls in the sixth transmembrane domain of ACKR3 (Fig. 1H), were not tested, mutations altering nearby charged residue D275 (as well as D179 and to a lesser extent K208) abolished CXCL12 binding. Additionally, mutations altering the cysteines involved in di-sulfide bond formation abolished CXCL11 and CXCL12 ligand binding, but mutations altering the N-terminal portion of the protein did not (see Fig. 1H). Although valine and methionine are both hydrophobic and thus the p.V258M substitution might be predicted to be tolerated, valine is C-beta branched that limits the conformations the main chain can adopt. By contrast, methionine contains a sulfur atom that can be involved in binding to metals (34). Thus, p.V258M substitution could alter slightly the conformation of the transmembrane domain and thus the binding pocket of the receptor, causing the observed decrease in binding affinity to the ligand. Moreover, a mild effect causing reduced function of the receptor is consistent with the human phenotype, as a more severe change altering the transmembrane domain would cause complete loss-of-function, which in the homozygous state in mouse causes severe phenotypes and high perinatal lethality (11, 12).

Ackr3^KO/KO embryos have misrouting of the oculomotor, abducens and motor trigeminal nerves, with significant variability not only between embryos but often between the two sides of the same embryo. The oculomotor axon misguidance, both in the midbrain and in the periphery, is similar to, but more variable than, what we reported with loss of Cxcr4 or Cxcl12 (4). In either Isl1-cre: Cxcr4^cKO/cKO:Isl^MN-GFP or Cxcl12^KO/KO:Isl^MN-GFP embryos, oculomotor axons grow dorsally rather than ventrally, stall and do not reach the orbit. The EOMs are then misinnervated by axons from the motor trigeminal nucleus that are aberrantly following the trigeminal sensory pathway. While in a few embryos there was unilateral formation of a thin nerve (4), there was significantly less variability than seen in the Ackr3^KO/KO embryos. Because ACKR3 sequesters CXCL12, loss of ACKR3 function results in more CXCL12 available for CXCR4 binding. We were initially unsure whether, in Ackr3^KO/KO mice, the loss of ACKR3-mediated regulation of CXCL12 levels would lead to increased or decreased CXCR4 signaling. Excess levels of CXCL12, however, lead to CXCR4 receptor downregulation and decreased CXCR4 signaling (8, 16, 19). Similar to reports in other brain regions (15, 16), we found that expression of CXCR4 protein (but not Cxcr4 mRNA) in and around the oculomotor nucleus was decreased with loss of Ackr3. This suggests that the increased levels of CXCL12 anticipated in the absence of ACKR3 result in increased CXCR4–CXCR12 binding followed by an increase in CXCR4 receptor internalization and downregulation. Consistent with this, in an ex vivo slice assay, addition of excess CXCL12 produces a strikingly similar phenotype to inhibition of CXCR4 with AMD3100. We hypothesize that the reduced binding affinity of V258M-ACKR3 lowers CXCR4 signaling just enough to effect ocular motor axon outgrowth.

Although the phenotype in mice from loss of Ackr3 is similar to loss of Cxcr4 or Cxcl12, it differs in significant ways, consistent with ACKR3’s multiple roles and binding partners. First, the oculomotor nucleus was uniformly affected with loss of Cxcr4 or Cxcl12. By contrast, in Ackr3^KO/KO embryos, axons from the caudal aspect of the nucleus always projected dorsally and aberrantly, but axons from the rostral portion of the nucleus often projected ventrally and formed a thin nerve that extended to the orbit. Interestingly, motor neurons in the rostral portion of the nucleus are born first, and their axons form the inferior division of CN3 to innervate the medial and inferior recti and the inferior oblique. The caudal motor neurons are born later and their axons form the superior division of CN3 (35, 36), to innervate the superior rectus muscle and levator palpabrae superioris. Thus, the selective vulnerability of the caudal axons is consistent with ptosis and secondary aberrant innervation of the levator palpebrae superioris by fibers of CN6, as seen in affected family members with synkinesis. Second, in Ackr3 knock-outs, some of the misprojecting axons project rostrally and cross the midline as early as E11.5. This was never observed with loss of Cxcr4 or Cxcl12. Normally, superior division cell bodies cross the midline beginning at E13.5, trailing their axons behind to project contralaterally. Slit/Robo signaling is involved in regulating this midline crossing, and premature cell body crossing is seen in mice lacking Slit/Robo signaling (37). Outgrowing axons do not normally cross the midline. The fact that this aberrant axonal crossing is seen only with loss of Ackr3, not Cxcr4 or Cxcl12, indicates that it likely results from a non-CXCR4 related pathway, such as binding to CXCL11. This remains an area for future study.

The loss of Ackr3 also has variable effects on CN6 development, ranging from its complete absence (with resulting Duane syndrome if CN3 axons are present), to normal innervation of the LR, to CN6 aberrantly innervating other EOMs. Interestingly, in the embryos with Duane syndrome pathology, all of the misinnervating fibers came from the oculomotor inferior decision region, in contrast to the MafB and Chn1 Duane mouse models, in which aberrant projections from CN3 to the LR come from both the superior and inferior decision regions (2, 3). This is likely because the embryos with aberrant innervation of the LR by CN3 also had thin oculomotor nerves with a relative paucity of superior division fibers. Interestingly, the half-sister of three of the affected individuals in this family (IV-1) was heterozygous for the ACKR3 mutation and had Duane syndrome, while the parents and some siblings of the affected, who are also heterozygous for the mutation, do not have any eye movement disorders. Thus, it is unresolved whether DRS represents variable penetrance of the heterozygous variant or if the etiology of her Duane syndrome was unrelated.

The existence of oculomotor synkinesis suggests that uninnervated EOMs secrete factors that attract motor neuron axons. This raises the question of whether CN6 would aberrantly innervate other EOMs in the absence of CN3, as CN3 does in the absence of CN6. In Wnt1 null mice, which lack CN3 and CN4, very sparse aberrant innervation of some EOMs was seen, but the source of those axons, presumed to be CN6, could not be traced to their origin (38). In Ackr3^KO/KO embryos we show CN6 contacting other EOMs in the orbit, but the aberrant CN6 branches do not extend to all muscles innervated by CN3. There may simply not be enough axons in CN6 to respond to all of the uninnervated muscles.

In zebrafish, knock-down of cxcr7b disrupted facial motor neuron migration from rhombomere 4 to rhombomere 6, leading to an accumulation of cell bodies in rhombomere 5 (18). In Ackr3^KO/KO mice, we also see disrupted facial motor neuron migration. Interestingly, we do not see any corresponding abnormalities of facial nerve trajectory or branching. This is similar to cdk5 knock-out mice, in which the facial motor neurons fail to migrate (39), but the facial nerve extends normally (A. Tenney, J. Livet, T. Belton, M. Prochazkova, E. Pearson, M. Whitman, A. Kulkarni, E. Engle, C. Henderson, manuscript in revision). The migration of the cell bodies, therefore, may not have functional significance for the development of the facial nerve. Why such a migration would have evolved is an open question.

One hallmark of the phenotype of Ackr3 knock-out mice is the high variability, ranging from gross malformations of the head to subtle nerve miswiring. This may be explained by variability in the amount of excess CXCL12 and resulting ligand-dependent downregulation of CXCR4 or by ACKR3’s role as a regulator of signaling pathways, which are also subject to other regulation. There have been reports of variable phenotypes with other CCDD genes as well. Patients with COL25A1 mutations have phenotypes ranging from unilateral ptosis to bilateral Duane syndrome (40). While CHN1 mutations typically cause Duane syndrome, they can also cause vertical strabismus and superior oblique palsy in the absence of Duane syndrome (41, 42). Thus, there may be an inherent variability in development of the ocular motor system.

The ability to accurately move one’s eyes and face is crucial to social communication. Synkinetic movements, which can be developmental or acquired after a transient CN palsy caused by aneurysms or tumors (43), can be significantly debilitating. Understanding the signaling pathways and axon guidance cues that pattern initial CN trajectories is a first step toward treatments to prevent or alleviate both developmental synkinesis and aberrant regeneration.

Materials and Methods

Genetic Methods

Families with complex ocular dysmotility disorders were enrolled in an ongoing research study at Boston Children’s Hospital following Institutional Review Board approval and informed consent. Detailed medical and family histories and clinical and neuroimaging data were obtained from participants and their medical records. Probands and their family members provided a blood and/or saliva sample from which genomic DNA was extracted using Gentra Puregene Blood Kits (Qiagen, Hilden, Germany) or Oragene DNA Purifier (DNA Genotek). Maternity and paternity were examined by the co-inheritance of at least 10 informative polymorphic markers.

SNP array and fluorescent polymorphic markers

An Affymetrix 10K SNP array was run on four affected individuals (IV-3, IV-9, IV-13, IV-14) and two unaffected parents (III-2 and III-3) and used for homozygosity mapping and multipoint linkage analysis. Fluorescent polymorphic markers in and near the regions of linkage were run on the four affected participants, two unaffected parents and seven unaffected siblings and were used for two-point linkage analysis.

Homozygosity mapping

Homozygosity mapper (homozygositymapper.org) (44) was used to identify homozygous regions. Regions in which the unaffected parents were homozygous for greater than five consecutive markers were excluded.

Linkage analysis

Multipoint linkage was calculated from the SNP array data using Linkdatagen (45) to generate input files for MERLIN (46) under a model of autosomal recessive inheritance with full penetrance. Two-point linkage analysis was calculated from the polymorphic markers using Fastlink (mlink) (47, 48) under a model of autosomal recessive inheritance with full penetrance. Loops in the pedigree were broken using the program unknown (49) prior to linkage analyses.

Next-generation sequencing

DNA from two affected individuals (IV-9 and IV-14) underwent whole-exome sequencing at the Broad Institute or Ocular Genomics Institute. DNA from a third affected individual (IV-13) underwent whole-genome sequencing at New York Genome Center.

Sanger validation

The identified variant was validated in all individuals by Sanger sequencing.

Analysis of variants

Next generation sequencing data were analyzed using seqr software (Broad Institute). Rare variants were defined as variants with allele frequency < 0.01 in the 1000 genomes, gnomAD, ExAC and TOPMed databases. Rare non-coding variants in the chromosome 2 linkage region in IV-13 that passed quality filters for allele frequency (>0.5) and read depth (>10) were examined using UCSC genome browser. Conservation at each location was determined based on 100 Vertebrates basewise conservation by PhyloP. Locations were deemed non-conserved if they had scores of <1.

ACKR3 reference sequence

All ACKR3 variants were based on GRCh37/hg19 and transcript isoform 1 (NM_020311.2) with nucleotide (cDNA) numbering using +1 as the A of the ATG translation initiation codon in the reference sequence, with the initiation codon as codon 1.

In silico analysis

The pathogenicity of the missense variant was assessed using in silico tools including CADD (Combined Annotation Dependent Depletion) score (25), Mutation Taster (21), SIFT (22), PolyPhen-2 (23) and Fathmm (24).

Functional methods

Generation of constructs

A plasmid containing wild-type human untagged ACKR3 cDNA was purchased from OriGene (SC112665). Site-directed mutagenesis to generate the c.772G>A variant was performed by GeneWhiz and confirmed by full sequencing of the resulting plasmid. Both constructs were then sub-cloned into the pSNAP plasmid (Cisbio) and confirmed by sequencing.

Transfection

HEK293T cells were purchased from ATCC (CRL-3216) and subcultured between 2 and 10 times prior to transfection. Cells were grown on tissue culture plates for the HTRF assay or glass coverslips coated with poly-d-lysine for immunocytochemistry. Plasmids encoding SNAP-tagged wild-type ACKR3, SNAP-tagged mutant ACKR3 or empty SNAP-tag were co-transfected with a GFP encoding plasmid using Lipofectamine 3000 (Life Technologies) or jetPRIME transfection reagent (Polyplus). Cells were harvested 24 h post-transfection.

Antibody staining

Coverslips with transfected cells were fixed with 4% paraformaldehyde, 4% sucrose for 30 min, then washed with PBS and blocked in 2% normal goat serum (blocking buffer) for 1 h. Mouse anti-human ACKR3 antibody (R&D systems, 1:500) was diluted in blocking buffer and applied for 2 h at room temperature. After three washes in PBS, goat anti-mouse Alexa 546 secondary antibody (1:1000 in blocking buffer, Life Technologies) was applied for 1 h. After washing in PBS and staining with DAPI, coverslips were mounted in Fluoromount-G (ThermoFisher Scientific) mounting media and imaged on a Nikon PerfectFocus microscope.

HTRF assay

Transfected cells were labeled with 100 nM SNAP Lumi4-Tb (Cisbio) in 1X Tag-lite labeling medium (TLB) at 37 C for 1 h, washed with 1X TLB and frozen in 5% DMSO in cell culture media at −80°C. All reactions were performed in triplicate in a 384-well plate. Two-fold serial dilutions of fluorescently labeled CXCL12 (Tag-lite CXCR4 Receptor Red Agonist, Cisbio, L0012RED) were prepared for final concentrations of 200 nM to 0.01 nM in 1X TLB. Cells were thawed at 37°C for 2 min and washed in TLB. A SpectraMax M5 plate reader was used to measure 620 nM fluorescence, and samples were diluted such that wild-type- and mutant-transfected cells had similar 620 nM fluorescence. A total of 10 ul cells were added to each well and the reaction was allowed to equilibrate for 3 h at room temperature. The SpectraMax M5 plate reader was used to measure 655 nM and 620 nM fluorescence. The ratio of 655/620 fluorescence was calculated for each well and normalized to background fluorescence and maximum binding for each individual experiment. Kd was calculated for each experiment, then averaged (N = 6 independent experiments, each in triplicate).

Mouse strains

Ackr3 floxed mice (Strain B6;129S-Ackr3^tm1Twb/Mmucd, stock No: 036715-UCD) were cryorecovered from MMRRC and crossed to Ella-cre (Jackson Laboratory: B6.FVB-Tg (EIIa-cre)C5379Lmgd/J, stock number 003724) for 2 generations to create full knock-outs. Ella-cre mice express cre recombinase under the control of the adenovirus Ella promoter that targets expression to the early mouse embryo, allowing germline deletion of floxed genes (50). Ackr3^+/− mice were crossed to transgenic ISL^MN:GFP reporter mice (MGI: J:132726; gift of Sam Pfaff) or to transgenic Hb9:GFP reporter mice (Jackson Laboratory: B6.Cg-Tg (Hlxb9-GFP)1Tmj/J, stock number: 005029). ISL^MN:GFP is not cytotoxic and specifically labels motor neurons with a farnesylated GFP that localizes to the membrane of cell bodies and axons, allowing for visualization of motor nerves during development (51). Hb9:GFP specifically GFP-labels motor neurons of CN6, CN12 and spinal motor neurons and is also not cytotoxic (52). All animal work was approved and performed in compliance with Boston Children’s Hospital Institutional Animal Care and Use Committee protocols.

In situ hybridization

Chromogenic in situ hybridization was performed using standard protocols (53) on 20 µm sections of fresh frozen tissue from both knock-out animals and littermate controls at E11.5 (n = 3) and from wild-type tissue at E13.5 (n = 3). Adjacent tissue sections were probed for Islet1 (to identify motor neurons) and Ackr3 or Cxcr4. The Ackr3 probe was based on that used by Genepaint (http://www.genepaint.org/results/ackr3) and covered residues 377-649 of mouse Ackr3 transcript variant 1 mRNA (NM_001271607.1). Primers: Ackr3: forward 5′GGACTCAAGGAGCAGGTCAC and reverse 5′T7seqCGTAGCCTGTGGTCTTAGCG. Cxcr4: forward 5′ATGGAACCGATCAGTGTGTGAGTA and reverse 5′T7seqATGCTCTCGAAGTCACATCCTT. Isl1 forward: CACTGTGGACATTACTCCCTCTTACAG and reverse: 5′T7seqGGAACATCTGAATGAATGTTCCTCATG.

Real-time PCR

Real-Time PCR (RT-PCR) was used to assess CXCR4 mRNA expression at E11.5. Oculomotor and trochlear nuclei were dissected from wild-type and Ackr3^−/− littermates. Total RNA was extracted using the RNeasy Plus Micro Kit (Qiagen) according to the manufacturer’s protocol. 25 ng of total RNA was reverse transcribed to cDNA using Maxima H minus First strand cDNA synthesis kit (Thermo Scientific), according to the manufacturer’s instructions. Real-time quantitative PCR was performed with the above prepared cDNA and iTaq Universal SYBR Green SuperMix (Biorad). The primers for CXCR4 and β-actin were as follows (54): Forward Primer (CXCR4) 5′-AGGAAACTGCTGGCTGAAAAGG-3′, Reverse primer (CXCR4) 5′-GGAATTGAAACACCACCATCCA-3′; Forward Primer (β-actin) 5′-GCATTGCTGACAGGATGCAG-3′, Reverse primer (β-actin) 5′-CCTGCTTGCTGATCCACATC-3′. The amplification conditions were 2 min at 95°C, followed by 40 cycles of 15 s at 95°C, 60 s at 60°C. Quantitation of mRNA was performed by using Applied Biosystem StepOnePlus Real-time PCR System (Applied Biosystem). The gene β-actin was used to normalize the mRNA levels of each sample (N = 5 wild type and 3 Ackr3^−/−). Statistics were calculated using GraphPad Prism.

Western blot

Western blot was used to quantify CXCR4 protein expression at E11.5. Whole brains were dissected from wild-type and Ackr3^−/− littermates and placed into RIPA buffer (Pierce) with protease and phosphatase inhibitors and homogenized by triturating. Centrifugation was used to remove cellular debris. A total of 20 µg protein were loaded on a 4–12% Bis-Tris gel (Novex by Life Technologies) and transferred to a nitrocellulose membrane. The blot was blocked with 5% milk in TBS with 1% Tween, incubated overnight in anti-CXCR4 antibody (1:1000, Abcam 124 824), washed and incubated in HRP-conjugated donkey-anti-rabbit (1:10000, Jackson ImmunoResearch). The blot was developed using ECL (GE Healthcare) and imaged on a LAS-4000 Luminescent Image Analyzer. The blot was then stripped and incubated in anti-beta actin HRP (C4) (1:40000, Santa Cruz Biotechnology) and developed and imaged. Images were quantified using ImageStudio software (N = 4 wild type and 6 Ackr3^−/−). Statistics were calculated using GraphPad Prism.

Fluorescent whole mount embryo immunohistochemistry: (A) BABB

Whole mount E11.5 or E12.5 embryos were prepared as previously described (55). Embryos were fixed overnight in 4% paraformaldehyde, dehydrated through a methanol series, fixed overnight in 1 part dimethyl sulfoxide: 4 parts methanol, then rehydrated through a methanol series. They were blocked in a solution of 5% goat serum and 20% DMSO in PBS for 3 h, placed in primary antibody diluted in blocking solution for 5 days, washed with blocking solution for 4 h, then placed in secondary antibody diluted in blocking solution for 2 days. Once stained, embryos were dehydrated through a methanol series and cleared in a solution of 2 parts benzoic acid:1 part benzyl benzoate. Cleared samples were mounted on coverslips with the benzoic acid-benzyl benzoate solution. (B) iDisco: Whole mount E13.5+ embryos were prepared using the iDISCO+ method as previously described (56), with longer (7 days) incubations in primary and secondary antibodies. Embryos were fixed overnight in 4% paraformaldehyde, dehydrated through a methanol series, bleached overnight in a 5% H₂O₂ in methanol, then rehydrated through a methanol series and washed in PBS-T (2% Triton-X in PBS). They were incubated in permeabilization solution for 2 days and then blocking solution for 2 days. They were then incubated with primary antibody diluted in a solution of PTwH (2% Tween-20 and heparin in PBS), 5% DMSO and 3% donkey serum for 7 days, washed with PTwH for 1 day, then placed in secondary diluted in a solution of PTwH with 3% donkey serum for 7 days. Once stained, embryos were washed in PTwH, dehydrated through a methanol series and incubated in a solution of dichloromethane and methanol. They were then incubated in dibenzyl ether to clear. Cleared samples were mounted on coverslips with dibenzyl ether.

Antibodies

Prepared embryos were stained with Mouse Anti-Actin, α-Smooth Muscle-Cy3™ (1:750; MilliporeSigma) and rabbit anti-GFP (1:500; Thermo Fisher Scientific) primary antibodies and Alexa-Fluor 647 goat anti-rabbit (1:1000; Invitrogen) secondary antibody. Anti-CXCR4 antibody (1:1000; Abcam 124824) was used for immunohistochemistry and western blot.

Imaging

Embryos were imaged on a Zeiss LSM 700 series laser scanning confocal microscope. Images were acquired using Zen Software (Carl Zeiss MicroImaging GmbH, Göttingen, Germany) and manipulated in three dimensions using Imaris software (Bitplane).

Quantification

The width of CN3 at E11.5 was quantified using Imaris software (Bitplane). The width of the nerve was measured at its origin where it first fully fasciculated after exiting the midbrain. Points were placed at the inferior and superior borders of the nerve and the distance in three dimensions between them calculated. GraphPad Prism 7 was used for statistical analysis [N = 16 wild-type nerves (from 8 embryos) and 24 knock-out nerves (from 12 embryos)].

Frozen section immunohistochemistry

Immunohistochemistry for CXCR4 and GFP was performed on 20 µm frozen sections. Antigen retrieval was performed using citrate-based buffer (Antigen Unmasking Solution H-3300, Vector Laboratories, Burlingame, CA) and steam for 20 min. Sections were then blocked in 2% normal donkey serum in 1% Triton-X in PBS, incubated in primary antibody overnight at 4°C, washed, incubated in secondary antibody 1 h at room temperature, washed and mounted in Flourogold mounting media. Images were acquired on a Revolve4 Fluorescent microscope (ECHO Laboratories), using identical settings (N = 3 for each condition).

Slice culture assay

The oculomotor ex vivo slice culture assay was performed as previously described (4). Briefly, embryos from Isl^MN:GFP mice were removed from the uterine horn at E10.5, embedded in 4% low-melting temperature agarose, oriented so the oculomotor nucleus and eye were in the same slice and 400 µm slices sliced on a vibratome in slicing buffer (HBSS (without Ca++ and Mg++) supplemented with HEPES and penicillin/streptomycin). The slices were then grown on millicell cell culture inserts with media (FlouroBright DMEM with 25% HBSS, 25% FBS, 0.5% glucose, 1 mM glutamine and 2.5 mM HEPES) underneath in six-well plates on a microscope stage incubator (37°C and 5% CO2). Fluorescent images were taken every 30 min for up to 2 days. Recombinant CXCL12 (R&D systems) was added to the media for a final concentration of 50 or 100 µM. These concentrations were chosen to fully saturate CXCR4 binding, as they are many folds higher than the reported Kd for CXCR4/CXCL12 binding (~4 nM) (57). The slice sits on a membrane above the media, and high concentrations allowed for the possibility of poor diffusion into, and therefore a lower effective concentration in the slice.

Funding

National Eye Institute (5K08EY027850, R01EY027421, R01EY15298); National Institutes of Health/Gabriella Miller Kids First Research Program (X01HL132377); National Institute of Child Health and Development (U54HD090255); Harvard-Vision Clinical Scientist Development Program (5K12EY016335); the Knights Templar Eye Foundation [Career Starter Grant]; and the Children’s Hospital Ophthalmology Foundation [Faculty Discovery Award]. N.M. was a Howard Hughes Medical Institute Medical Research Fellow; P.M.M.R. was a Howard Hughes Medical Institute Exceptional Research Opportunities Program Student; and E.C.E. is a Howard Hughes Medical Institute Investigator.

Conflict of Interest Statement

None declared.