Inter-axonal molecular crosstalk via Lumican proteoglycan sculpts murine cervical corticospinal innervation by distinct subpopulations

Abstract

How CNS circuits sculpt their axonal arbors into spatially and functionally organized domains is not well understood. Segmental specificity of corticospinal connectivity is an exemplar for such regional specificity of many axon projections. Corticospinal neurons (CSN) innervate spinal and brainstem targets with segmental precision, controlling voluntary movement. Multiple molecularly distinct CSN subpopulations innervate the cervical cord for evolutionarily enhanced precision of forelimb movement. Evolutionarily newer CSN_BC-lat exclusively innervate bulbar-cervical targets, while CSN_medial are heterogeneous; distinct subpopulations extend axons to either bulbar-cervical or thoraco-lumbar segments. We identify that Lumican controls balance of cervical innervation between CSN_BC-lat and CSN_medial axons during development, which is maintained into maturity. Lumican, an extracellular proteoglycan expressed by CSN_BC-lat, non-cell-autonomously suppresses cervical collateralization by multiple CSN_medial subpopulations. This inter-axonal molecular crosstalk between CSN subpopulations controls murine corticospinal circuitry refinement and forelimb dexterity. Such crosstalk is generalizable beyond the corticospinal system for evolutionary incorporation of new neuron populations into preexisting circuitry.

Keywords: CP: Neuroscience; Lumican proteoglycan; axon collateralization; axon development; circuit refinement; corticospinal neurons; corticospinal segmental specificity; evolutionary circuit diversification; forelimb dexterity; inter-axonal molecular crosstalk.

Figure 1.. Lumican is expressed by CSN_BC-lat in postnatal developing cortex in mice

(A) Schematized representation of developmentally distinct CSN subpopulations. CSN in rostrolateral sensorimotor cortex (CSN_BC-lat; green) and CSN in caudomedial sensorimotor cortex (CSN_medial; red) are illustrated. For simplicity, axon collaterals are not illustrated, and CSN are illustrated in only one hemisphere. (B) Temporal profile of Lumican expression from microarray analysis. CSN_BC-lat, green; CSN_medial, red. The y axis represents normalized expression level; data are presented as the mean ± SD, n = 2–3. (C–E) In situ hybridization on coronal brain sections reveals that Lumican is expressed exclusively in lateral cortex. Scale bars, 400 μm. (F–H) Immunocytochemistry on coronal P4 brain section detects Lumican expression (green) only in lateral cortex, while it is not detected medially (F, F′). Lumican is expressed specifically in layer V (G, G′), and virtually all Lumican-positive cells co-express CTIP2 (magenta) (H, H′). Arrowheads indicate neurons co-expressing Lumican and CTIP2. Scale bars, 500 μm (F); 100 μm (G); 50 μm (H). (I) There is no Lumican expression (green) in P7 Fezf2 null cortex, while meningeal Lumican expression is preserved. Arrowheads indicate layer V neurons expressing CTIP2 and Lumican in wild-type cortex, which are absent in Fezf2 null cortex. CTIP2, magenta; DAPI, blue. Scale bar, 500 μm. (J) Lumican (green) immunocytochemistry on coronal section of P5 brain retrogradely labeled by CTB-555 (magenta) injection into the cervical dorsal funiculus at P4. Co-labeling by Lumican and CTB is observed predominantly in lateral cortex, demonstrating Lumican expression selectively by CSN_BC-lat. Mean ± SD, n = 3. Scale bars, 500 μm. (K) Schematized representation of Lumican expression by developing CSN_BC-lat (green) between P4 and P14. CC, corpus callosum; M1, primary motor cortex; M2, secondary motor cortex; Str, striatum. See also Figure S1.

Figure 2.. Lumican suppresses CSN_medial axon collateralization in the cervical cord in a non-cell-autonomous manner

(A) Schematic illustrating the experimental outline for (B–I). (B and C) Representative coronal brain sections showing BDA injection site in M1. BDA, black. Scale bar, 1 mm. (D and E) Axial sections of cervical cord (C1–C2 level) show corticospinal axons entering the spinal cord. CST axons in the dorsal funiculus are enlarged in magnified images (D′ and E′). Note exuberant axon collateralization (white arrowheads) in Lumican null cervical gray matter, quite rare in wild type. BDA, black. Scale bars, 500 μm (D, E); 50 μm (D′, E′). (F and G) Horizontal sections of cervical cord show collateral innervation into the gray matter. Manually traced axon collaterals are labeled in red. Midline and outer border of gray matter are labeled in blue and green, respectively (F′, G′). BDA, black. Scale bar, 200 μm. (H) Manually traced axon collaterals (red) on serial horizontal sections (C3–C8) are collapsed onto a single plane image, together with midline (blue) and gray matter border (green). Scale bar, 500 μm. (I) Quantification of relative corticospinal axon collateral density in the cervical gray matter normalized to CST axon number at C1–C2 shows increased CSN_medial axon collateralization in Lumican null mice compared with wild-type mice. Mean ± SD, n = 3, Student’s t test. WT, wild-type; KO, Lumican null. (J) Schematic illustrates the experimental outline. AAVs encoding Cre recombinase (at a lower titer) and Cre-dependent tdTomato (at a higher titer) are stereotactically injected into rostrolateral sensorimotor cortex. Axial sections of C1–C2 cervical cord display tdTomato fluorescence in the dorsal funiculus and gray matter. Dashed lines demarcate gray matter. Quantification of relative density of CSN_BC-lat axon collaterals normalized to CST axon number at C1–C2 does not show a significant change in Lumican null compared with wild-type mice, although there appears to be a trend toward a modest reduction in Lumican null mice (see discussion). Mean ± SEM, n = 8 (wild type) or 5 (Lumican null), Student’s t test. Scale bar, 200 μm. See also Figure S2.

Figure 3.. CSN specification and axon extension are unchanged in Lumican null mice

(A) In the absence of Lumican, overall brain structure remains unchanged. SCPN, CPN, and CThPN marker and control molecules CTIP2, SATB2, and TBR1, respectively, are expressed and positioned normally at P8. Quantification of the number of neurons expressing these molecules in Lumican null primary somatosensory cortex is presented as a percentage of wild-type neurons. Mean ± SEM, n = 4 (wild type) or 3 (Lumican null). Scale bars, 200 μm. (B) Expression of CSN_BC-lat-specific genes Klhl14 and Cartpt and CSN_medial-specific genes Cry-mu and Crim1 is essentially indistinguishable between wild-type and Lumican null cortex at P4. Scale bar, 200 μm. (C) Total number and distribution of retrogradely labeled CSN are indistinguishable between wild-type and Lumican null cortex. Mean ± SEM, n = 3. Scale bar, 1 mm. (D) CST axons are anterogradely labeled by Emx1^{IRES-Cre/IRES-Flpo};Ai65^RCFL-tdT/+. Lumican null CST axons normally decussate in the medulla and extend in the spinal cord. Mean ± SEM, n = 4. Scale bars, 100 μm.

Figure 4.. Lumican overexpression by CSN_BC-lat non-cell-autonomously suppresses CSN_medial axon collateralization in the cervical cord

(A) Schematic illustrating the experimental outline. (B–E) Axial sections of C1–C2 cervical cord show corticospinal axons labeled by AAV-derived EGFP (green; B and C) or BDA (black; D and E). Scale bars, 500 m. (F and G) Horizontal sections of cervical cord (C3– C8) display CSN_medial axon collateral innervation into the gray matter. BDA, black. Scale bars, 500 μm (F, G); 200 μm (F′, G′). (H) Quantification of relative corticospinal axon collateral density normalized to CST axon number at C1–C2 shows that Lumican overexpression by CSN_BC-lat substantially reduces CSN_medial axon collateralization. Mean ± SD, n = 3, Student’s t test. See also Figures S3 and S4.

Figure 5.. Exclusionary, subtractive viral labeling reveals increased CSN_BC-med axon collateralization in Lumican null cervical cord

(A) Schematized representation of developmentally distinct CSN subpopulations by the combination of location and projection specificity. Compared with the schematic in Figure 1A, CSN in caudomedial sensorimotor cortex (CSN_medial) are further subdivided into CSN_BC-med (blue) and CSN_TL (orange). While CSN_BC-med send projections only to brainstem and cervical cord, CSN_TL axons extend and collateralize throughout the spinal cord. For simplicity, axon collaterals are not illustrated, and CSN are illustrated in only one hemisphere. (B) Schematic illustrating the experimental outline. (C and D) Axial sections of C3–C4 cervical cord show turboRFP and EGFP fluorescence in the dorsal funiculus and gray matter. In higher-magnification images (C_1–4, D_1–4), magenta arrowheads indicate turboRFP⁺;EGFP⁻ axon collaterals, while yellow arrowheads indicate turboRFP⁺;EGFP⁺ axon collaterals. To identify turboRFP⁺;EGFP⁻ collateral signal, turboRFP⁺ and EGFP⁺ pixels above threshold are separately identified, followed by subtraction of EGFP⁺ pixels from turboRFP channel (C₄, D₄). Scale bars, 200 μm (C, D); 50 μm (C_1–4, D_1–4). (E) Quantification of relative density of turboRFP⁺;EGFP⁻ axon collateral density normalized to CST axon number at C1–C2. Mean ± SEM, n = 3, Student’s t test. See also Figure S5.

Figure 6.. Intersectional genetic labeling reveals increased CSN_TL axon collateralization in Lumican null cervical cord in 5-week-old mice, but not at P15

(A) Schematic illustrating the experimental outline. (B and C) Immunocytochemistry on axial sections of C3–C4 cervical cord at 5 weeks of age display tdTomato⁺ corticospinal axons labeled by tamoxifen injection at P3 into Lumican wild-type or null Crim1^GCE/+;Emx1^{IRES-Flpo/IRES-Flpo};Ai65^{RCFL-tdT/RCFL-tdT} mice. Dashed lines demarcate gray matter. Scale bar, 500 μm. (D) Quantification of relative corticospinal axon collateral density normalized to CST axon number at C1–C2 shows increased CSN_TL axon collaterals in Lumican null mice compared with wild-type mice throughout the cervical cord. Mean ± SEM, n = 7, Student’s t test. (E and F) Distribution of CSN_TL axon collaterals was quantified along mediolateral (E) and dorsoventral (F) axes in wild-type and Lumican null cervical gray matter at the C3–C4 level. A representative image is shown to delineate cervical gray matter (dashed line). Mean ± SEM, n = 7, Student’s t test. Statistical analysis was performed in each of four bins along each axis. Scale bar, 200 μm. (G) Immunocytochemistry on axial sections of T1–T2 thoracic cord and L1–L2 lumbar cord collected at 5 weeks of age shows tdTomato⁺ corticospinal axons labeled by tamoxifen injection at P3. Dashed lines demarcate gray matter. Scale bar, 500 μm. (H) Quantification of CSN_TL axon numbers in the dorsal funiculus white matter shows indistinguishable rostral-to-caudal reduction in wild-type and Lumican null mice. Mean ± SEM, n = 4, Student’s t test. (I) Quantification of relative CSN_TL axon collateral density at T1–T2 and L1–L2 normalized to CST axon number at each respective segment shows indistinguishable CSN_TL axon collateral density in Lumican null mice compared with wild-type mice. To enable direct comparison, the data for C1–C2 from (D) are reproduced here at this scale. Mean ± SEM, n = 7, Student’s t test. (J) Immunocytochemistry on axial sections of C1–C2 cervical cord at P15 displays tdTomato⁺ corticospinal axons labeled by tamoxifen injection at P3 into wild-type or Lumican null Crim1^GCE/+;Emx1^{IRES-Flpo/IRES-Flpo};Ai65^{RCFL-tdT/RCFL-tdT} mice. Dashed lines demarcate gray matter. Scale bar, 200 μm. (K) Quantification of relative corticospinal axon collateral density normalized to CST axon number at C1–C2 shows indistinguishable CSN_TL axon collaterals in P15 Lumican null mice compared with wild-type mice at C1–C2 and C3–C4. Mean ± SEM, n = 6 (wild type) or 5 (Lumican null), Student’s t test. See also Figure S6.

Figure 7.. Reduced forelimb dexterity in Lumican null mice and expression of SLRP family genes in marmoset and mouse brains suggest expanded use of SLRP genes in mammalian CSN evolution for potentially enhanced forelimb dexterity

(A) Quantification of mean velocity and total distance moved in open field test shows normal gross locomotor activity of Lumican null mice. Mean ± SEM, n = 10 (wild type) or 17 (Lumican null), Student’s t test. (B and C) A timeline of the pellet grasping test (B). Success rates of pellet grasping during training period show significantly reduced success rate in Lumican null mice compared with wild-type mice (C). Mean ± SEM, n = 16 (wild type) or 14 (Lumican null), two-way repeated measures ANOVA. (D) Schematic illustrating inter-axonal molecular crosstalk between molecularly distinct CSN subpopulations. Lumican is expressed by evolutionarily newer CSN_BC-lat (upper inset; green), non-cell-autonomously sculpting axon collateralization by evolutionarily older CSN_BC-med (blue) and CSN_TL (orange) in cervical spinal cord (lower inset). CSN_BC-med and CSN_TL are interdigitated in medial cortex; axons of all three subpopulations are interdigitated in the CST. Primary (M1) and secondary (M2) motor, and primary somatosensory (S1), cortices are outlined in the right cortex (outlines adapted from Kaas). (E) Nissl staining and in situ hybridization for CSN/SCPN control genes FEZF2 and CTIP2/BCL11B, CSN subpopulation-specific genes (FRZB for CSN_BC-lat and CRIM1 for CSN_medial/CSN_TL, see Sahni et al.), and SLRP family genes LUMICAN and DECORIN on P0/P1 marmoset coronal brain sections. All images shown here are taken from the Marmoset Gene Atlas website (https://gene-atlas.bminds.brain.riken.jp; experiments AI-1, 78–2, 140–4, 251–7, 262–5, 301–8). Scale bar, 1 mm. (F) Temporal profile of SLRP family gene expression from microarray analysis. CSN_BC-lat, green; CSN_medial, red. Lumican graph is also shown in Figure 1B. Other than Lumican, Fibromodulin is the only SLRP gene showing any significant differential expression: Fibromodulin expression peaks around P4, although at a much lower level than Lumican. Decorin does not show any significant, detectable expression. The y axis represents normalized fluorescence intensity; mean ± SD, n = 2–3. (G) Immunocytochemistry on coronal section of P8 wild-type mouse brain showing Decorin (magenta) and CTIP2 (green). Apart from meningeal expression, Decorin is not expressed in rostral cortex. Scale bars, 500 μm. See also Figure S7.