Knockout of Celsr3 in the brainstem impairs development of the RST
Our previous studies validated Cre-loxP induced inactivation of “floxed” Celsr3 gene [15–18, 13]. To indirectly monitor Celsr3 inactivation upon En1-Cre expression, we generated En1-Cre;Rosa26GFP mice. In sagittal sections of embryonic day (E) 13.5 brain, GFP signal was concentrated in the anlage of brainstem and cerebellum (Fig. 1A, n = 3). In the cerebellum, Celsr3 is expressed in Purkinje cells and Celsr3 knockout affects motor learning but not walking ability in our recent report . Here, we focused on inactivation of Celsr3 in the brainstem. En1-Cre;Celsr3f/− (Celsr3 cKO) mice survived well and looked similar to littermate controls (Fig. 1B).
As En1-Cre is strongly expressed in the midbrain, we first studied RST projections in Celsr3 cKO animals using anterograde and retrograde tracing. FG was injected into right C6 spinal segment and retrogradely labelled neurons were counted in the left midbrain (Fig. 1C, D).AAV9-GFP was injected into left red nuclei and the anterogradely labelled fibers were observed in right halves of spinal cords (Fig. 1C, E). In serial parasagittal sections, FG-labelled red nuclei were readily identified in the midbrain in both control and mutant mice, but their number was dramatically reduced in the mutant (Fig. 1F1-F6, G1-G6). The total labelled red nuclei neurons were reduced by 83% in Celsr3 cKO compared with the control (Fig. 1H; control and mutant: 667 ± 22 and 113 ± 7, P < 0.0001, Student’s t-test, n = 6 in each group).
Four weeks after AAV9-GFP injection into red nuclei (Fig. 1C, E), RST axons were visible in the superficial layer of dorsolateral white matter, and scattered in the gray matter in the control (Fig. 1I, n = 3), whereas rare GFP-labelled axons could be identified in mutant spinal segments (Fig. 1J, n = 3). Thus, in line with previous studies in other brain regions [16, 13], Celsr3 regulates the development of rubrospinal projections in a cell-autonomous manner.
Celsr3 knockout in the brainstem partially affects CST projections but not ascending trigeminal lemniscus
The brainstem is an important intermediate target for the pathfinding of corticospinal axons. To assess whether Celsr3 inactivation impairs the development of the CST, we studied axon projections using anterograde tracing by injecting AAV9-hSyn-GFP virus into the motor cortex and retrograde tracing by injecting FG into C6 spinal segment (Fig. 2A). GFP-labelled CST axons were well visualized to travel through the brainstem in both groups, but some axons stopped and misrouted in the ventral region of the midbrain in Celsr3 cKO mice (Fig. 2B). In FG tracing studies, retrogradely labelled neurons were found in layer V of contralateral parasagittal hemisphere sections, and displayed a laminar organization in the deep layer of the cortex (Fig. 2C), and their numbers were significantly reduced in the mutant compared to the control (Fig. 2D; control and mutant in cells/section: 201 ± 5 and 161 ± 11, P = 0.0168, Student’s t-test, n = 4 and 3 in the control and mutant respectively). In addition, we carried out anti-PKCγ immunostaining with C5-T1 spinal segments. In agreement with a previous report , PKCγ immunoreactivity disclosed CST descending axons in the dorsal funiculus in two groups, but axon bundles were disorganized in Celsr3 cKO mice (Fig. 2E). The fiber density was decreased by 30% in the mutant compared to the control (Fig. 2F; control and mutant in 107 pixels/section: 2.98 ± 0.22 and 2.08 ± 0.12, P = 0.0108, Student’s t-test, n = 4 in each group). These results show that Celsr3 in the brainstem regulates the development of the CST, in a non-cell autonomous manner.
The relay nuclei of trigeminal afferent pathways are located in the brainstem, and the pathway can be readily assessed by examining the organization of barrels in sensory cortex . Adult hemispheres were prepared as flattened mounts and stained with anti-vGlut2. In primary somatosensory cortex, cortical barrels were identical in the mutant and the control (Fig. 2G), indicating that the development of the ascending sensory tract is not primarily dependent on Celsr3 expression in the brainstem.
Celsr3 cKO mice show increased branching of dopaminergic fibers in spinal segments
Projections of dopaminergic axons from the diencephalon and serotonergic fibers from the raphe progress through the brainstem and innervate different spinal segments. Both are involved in motor control. We carried out anti-TH and − 5-HT immunostaining of transverse spinal sections at C8-T1 segments and, performed tridimensional reconstruction of fibers (Fig. 3A-H). Intriguingly, in Celsr3 cKO animals, the density of dopaminergic fibers was increased in the ventral horn (Fig. 3C, D), but not in the intermediate zone (Fig. 3A, B), compared to control animals (Fig. 3I; control and mutant in µm/section: 9222 ± 457 and 11276 ± 336 in the ventral horn, 11646 ± 1672 and 13241 ± 549 in the intermediate zone, P = 0.0223 and 0.4161 respectively, Student’s t-test, n = 3 in each group). On the other hand, the density of serotoninergic fibers was comparable in both groups (Fig. 3E-H, J; control and mutant in µm/section: 8090 ± 591 and 9072 ± 574 in the ventral horn, 8169 ± 1116 and 10123 ± 278 in the intermediate zone, P = 0.2991 and 0.644 respectively, Student’s t-test, n = 3 in each group).
Celsr3 is known to regulate dopaminergic axon pathfinding in a cell-autonomous manner , and dopaminergic neurons projecting to spinal cord are mainly derived from diencephalic A11 region [23–25]. To try and understand the dopaminergic fiber phenotype in spinal segments of Celsr3 cKO mutants, we studied Cre expression in dopaminergic neurons using En1-Cre;Rosa26GFP mice combined with anti-TH immunostaining, and found no double-labelled neurons in the A11 region (Fig. 3K). This indicates that Celsr3 expression is preserved in spinal-projecting dopaminergic neurons in Celsr3 cKO mice. Using FG retrograde labeling combined with TH immunostaining of sagittal brain sections, FG-labelled neurons in A11 were positive for TH and present in comparable numbers in mutant and control samples (Fig. 3L, M; control and mutant in cells/section: 55 ± 17 and 45 ± 19, P = 0.7216, Student’s t-test, n = 3 in each group). These results suggest that increased dopaminergic fiber density in Celsr3 cKO animals is due to compensatory axonal branching when the CST and RST are defective.
Maturation and output of spinal motoneurons are impaired in Celsr3 cKO mice
Corticospinal, rubrospinal and propriospinal projections make direct and indirect connections with spinal motoneurons, and defects of these descending tracts may impact motoneuron maturation and function. To test this, transverse spinal sections at C5-T1 were prepared for anti-ChAT immunostaining (Fig. 4A). The number of spinal motoneurons was subtly decreased in the mutant compared to the control (Fig. 4B; control and mutant in cells/section: 56.5 ± 0.4 and 50.1 ± 0.7, P < 0.0001, Student’s t-test, n = 5 in each group). Upon stimulation of musculocutaneous nerves, the EMG recording of biceps showed a significant reduction in amplitude, but not of latency, in Celsr3 cKO mice compared to control mice (Fig. 4C-E; control and mutant: 17.3 ± 0.7 and 10.5 ± 1.1 mV in amplitude, 0.79 ± 0.03 and 0.83 ± 0.06 ms in latency; P = 0.0001 and 0.6066 respectively, Student’s t-test, n = 8 in each group). The mutant biceps were hypotrophic, with a significant reduction of wet weight compared to controls (Fig. 4F, G; control and mutant in mg: 27.8 ± 2.6 and 19.4 ± 1.1, P = 0.0097, Student’s t-test, n = 4 and 6 in the control and the mutant). We studied NMJs by staining for NF200 (fiber terminus) and α-BT (acetylcholine receptors) in biceps, and found a 33% of reduction in the mutant compared to the control (Fig. 4H, I; control and mutant in NMJs/muscle: 743 ± 26 and 499 ± 28, P = 0.0003, Student’s t-test, n = 4 and 6 in the control and the mutant).
Celsr3 cKO mice have defective motor coordination but normal walking ability and fine-movement control
As described above, various descending axonal tracts were modified in Celsr3 cKO mice. To assess the consequences in terms of motor behavior, we studied locomotion and fine movement control. In open-field tests, there was no significant differences of movement trajectory and total distance between control and mutant mice (Fig. 5A, B; control and mutant in m: 94.8 ± 4.8 and 97.8 ± 5.0, P = 0.6802, Student’s t-test, n = 8 and 6 in the control and the mutant). The grip strength of forelimbs was similar in both groups (Fig. 5C; control and mutant in gf: 111.2 ± 3.9 and 110.8 ± 4.5, P = 0.9376, Student’s t-test, n = 8 and 6 in the control and the mutant). In contrast, the percentage of footslips in grid tests was significantly increased in Celsr3 cKO compared to controls (Fig. 5D, E; control and mutant: 5.33 ± 0.99% and 21.00 ± 3.96%, P = 0.0033, Student’s t-test, n = 6 in each group). In the Rotarod test, Celsr3 cKO mice had a shorter falling latency than control mice (Fig. 5F; control and mutant in sec: 183.3 ± 12.5 and 105.5 ± 14.9, P = 0.0003, Student’s t-test, n = 22 and 15 in the control and the mutant). Unexpectedly, the IBB scores in food-pellet taking were comparable in both groups (Fig. 5G, H; control and mutant: 8.91 ± 0.09 and 8.07 ± 0.53, P = 0.055, Student’s t-test, n = 23 and 14 in the control and the mutant), indicating that skilled movement is not affected in the mutant. The voluntary gait, assessed using the Catwalk (Fig. 5I), showed a longer forepaw stride (Fig. 5J; control and mutant in cm: 6.37 ± 0.18 and 7.41 ± 0.19, P = 0.0004, Student’s t-test, n = 22 and 15 in the control and the mutant), and a significant increase of forepaw swing in Celsr3 cKO compared to control mice (Fig. 5K; control and mutant in sec: 0.099 ± 0.005 and 0.118 ± 0.007, P = 0.0215, Student’s t-test, n = 23 and 14 in the control and the mutant).
Response to strong mechanical stimulation is impaired in Celsr3 cKO mice
Recent studies showed that descending projections from the cortex modulate sensory processing in spinal cord . We wondered whether any sensory abnormalities are present in Celsr3 cKO mutants, and carried out pain, thermal and mechanical sensation tests in adult animals. Interestingly, in von Frey tests, the percentage of hindpaw withdrawal was significantly decreased with stimuli at 0.4 g and 0.6 g stiffness in Celsr3 cKO mice compared to control mice (Fig. 6A; control and mutant: 90 ± 4% and 71 ± 4% at 0.4 g, 97 ± 3 and 73 ± 3 at 0.6 g, P = 0.0064 and 0.0002, Student’s t-test, n = 6 and 7 in the control and the mutant), indicating an impairment of mechanical sensation in the mutant. Upon laser stimulation of paws, the withdrawal latencies of forepaws and hindpaws showed no differences between both groups (Fig. 6B-D; control and mutant: 51.7 ± 0.6 and 51.6 ± 0.9°C for hot plate, 2.24 ± 0.07 and 2.23 ± 0.07 sec for forepaws, 2.86 ± 0.15 sec and 2.81 ± 0.16 sec for hindpaws, P = 0.9124, 0.9242 and 0.8392, Student’s t-test, n = 6 and 7 in the control and the mutant).
In order to know whether the RST is involved in the modulation of nociceptive mechanical perception, we studied the synaptic connections of rubrospinal axons with spinal neurons by injecting transsynaptic scAAV1-hSyn-Cre virus into the red nuclei of Ai14-tdTomato mice (Fig. 6E), and found that Tomato-positive spinal neurons were scattered in layer V-VII of the contralateral gray matter (Fig. 6F), suggesting that rubrospinal axons directly synapse on spinal neurons in the deep layer of the dorsal horn. Upon mechanical stimuli on hindpaws, calcium activities of contralateral red nuclei were recorded (Fig. 6G, H). The calcium signal in red nuclei was strong in controls (Fig. 6I, K), but less prominent in Celsr3 cKO mutants (Figure J, L). Quantitative analysis showed a significant decrease of calcium signal peak in the mutant compared to the control (Fig. 6M; control and mutant in ΔF/F: 12.39 ± 1.03% and 6.47 ± 0.53%; P = 0.007, Student’s t-test, n = 3). This suggests that the RST is required for the response to mechanical stimulation.