Crim1 and Kelch-like 14 exert complementary dual-directional developmental control over segmentally specific corticospinal axon projection targeting.

The cerebral cortex executes highly skilled movement, necessitating that it connects accurately with specific brainstem and spinal motor circuitry. Corticospinal neurons (CSN) must correctly target specific spinal segments, but the basis for this targeting remains unknown. In the accompanying report, we show that segmentally distinct CSN subpopulations are molecularly distinct from early development, identifying candidate molecular controls over segmentally specific axon targeting. Here, we functionally investigate two of these candidate molecular controls, Crim1 and Kelch-like 14 (Klhl14), identifying their critical roles in directing CSN axons to appropriate spinal segmental levels in the white matter prior to axon collateralization. Crim1 and Klhl14 are specifically expressed by distinct CSN subpopulations and regulate their differental white matter projection targeting-Crim1 directs thoracolumbar axon extension, while Klhl14 limits axon extension to bulbar-cervical segments. These molecular regulators of descending spinal projections constitute the first stages of a dual-directional set of complementary controls over CSN diversity for segmentally and functionally distinct circuitry.

INTRODUCTION

The corticospinal system is the principal spinal-projecting pathway controlling performance of highly skilled and complex movements (Martin, 2005). In addition, the corticospinal circuit exerts critical control over sensory modulation and autonomic functions (Lemon, 2008; Lemon and Griffiths, 2005; Liu et al., 2018; Sahni et al., 2020; Welniarz et al., 2017). For this precise top-down sensorimotor and autonomic control, distinct corticospinal neurons (CSN) extend axons to, and innervate, distinct subcerebral targets: rostral targets in the brainstem and cervical cord to caudal targets in the thoracic and lumbar cord. The molecular basis for this segmentally specific connectivity remains unknown.

Multiple investigations in mature CSN circuitry have aimed at understanding this diversity to elucidate functional diversity of cortical control over movement (e.g., CSN projecting to cervical cord broadly control arm and finger movement, while CSN projecting to lumbar cord broadly control leg movement). The somatotopic organization of adult cortex, wherein cortical regions are systematically organized to control movement of distinct body parts, has been well documented in the field, including in humans (Penfield and Boldrey, 1937; Penfield and Rasmussen, 1950), non-human primates (Woolsey et al., 1952), and rodents (Akintunde and Buxton, 1992; Li et al., 1990; Miller, 1987; Tennant et al., 2011; Ullan and Artieda, 1981; Wise et al., 1979; Donoghue and Wise, 1982; Neafsey et al., 1986).

Rodent studies have identified that CSN residing in distinct cortical locations, including outside primary motor cortex (M1), exhibit distinct spinal connectivity (Ueno et al., 2018; Tennant et al., 2011; Wang et al., 2017). The cortical area extending projections to lumbar cord is significantly smaller than cortical areas originating corticospinal projections to cervical cord (Kamiyama et al., 2015; Wang et al., 2018). Understanding how corticospinal outputs target specific spinal segments is a major goal toward identifying the developmental basis for the organization of descending neocortical control over distinct, segmentally specific, function—including motor control, sensory modulation, and autonomic control.

Transcriptional regulators have been identified as controlling the identity of broad neocortical projection neuron subtypes (Fame et al., 2011; Franco and Müller, 2013; Greig et al., 2013; Leone et al., 2008; Lodato et al., 2015; Molyneaux et al., 2007). In particular, specification and differentiation of CSN is regulated first by controls over broad corticofugal projection neuron differentiation, then by controls over all subcerebral projection neurons, and then by CSN-subtype-specific molecular controls (Arlotta et al., 2005; Cederquist et al., 2013; Chen et al., 2005a, 2005b; Galazo et al., 2016; Greig et al., 2016; Han et al., 2011; Joshi et al., 2008; Kwan et al., 2008; Lai et al., 2008; Lodato et al., 2014; McKenna et al., 2011; Molyneaux et al., 2005; Ozdinler and Macklis, 2006; Shim et al., 2012; Tomassy et al., 2010; Woodworth et al., 2012, 2016). Guidance molecules direct CSN axons at specific choice points, including in the spinal cord (e.g., EphA4; reviewed by Canty and Murphy 2008; Sahni et al. 2020; Welniarz et al. 2017). Further, specific signaling mechanisms promote CSN axon extension into the cord (e.g., insulin growth factor [IGF]-I [Ozdinler and Macklis, 2006] and Wnt-Ryk signaling [Liu et al., 2005]). However, all these mechanisms, including CSN-specific molecular controls identified to date, do not distinguish between hodologically distinct CSN subpopulations, and the molecular basis for the segmentally specific CSN connectivity described above remains entirely unknown.

In the accompanying study (Sahni et al., 2021), we identify that developing CSN subpopulations exhibit striking axon-targeting specificity in spinal white matter, and this establishes the foundation for durable specificity of adult corticospinal circuitry. Based on this developmentally specified axon targeting specificity, we are able to define segmentally distinct CSN subpopulations through development into maturity: (1) CSN in lateral cortex that reside outside ‘‘classic’’ motor cortex and project only to bulbar-cervical targets (CSNBC-lat); (2) CSN in medial sensorimotor cortex that project only to cervical segments and not beyond (CSNBC-med); and (3) CSN that extend axons past thoracic T2 into caudal thoracic and lumbar spinal segments (CSNTL) (Figure 1A). We also identified genes differentially expressed between bulbar-cervical (CSNBC) and thoracolumbar (CSNTL) projecting CSN subpopulations at critical developmental times. We identified that these segmentally distinct CSN subpopulations are molecularly distinct from the earliest stages of axon extension, and this molecular delineation goes beyond spatial separation in cortex.

Figure 1.

CSN are specified normally and project normally to the spinal cord in Crim1-null mice

(A) Schematic outlining delineation of three segmentally specific CSN subpopulations (spatially distinct CSNBC-lat and interdigitated CSNBC-med and CSNTL), indicating their molecular expression, axon extension, and axon collateralization, from the accompanying paper (Sahni et al., 2021). These CSN subpopulations are molecularly delineated during development and display persistent axon targeting before their final connectivity is established. Klhl14 expression delineates Klhl14-positive CSNBC-lat (green) from Klhl14-negative CSNBC-med (purple); all CSNTL (red) are Klhl14 negative, and 95% ± 2% of CSNTL express Crim1 during early development.

(B) Overlay of inverted and pseudocolored in situ hybridization images from the accompanying paper (Sahni et al., 2021) showing complementary expression of Klhl14 (green) and Crim1 (red) in cortex. The inset (Bʹ) shows Klhl14 is restricted to CSNBC-lat in lateral cortex.

(C–E′′) In situ hybridization from P4 Crim1 WT, heterozygous, and homozygous mice. Homozygotes are null for Crim1 expression (C–C′′) but exhibit normal Fezf2 (D′′; all CSN) and Cry-mu (E′′; CSNmedial) expression.

(F-I′′) Coronal sections of P4 brains retrogradely labeled from C1 at P3. Crim1 nulls have normal CSN distribution.

(J) Quantification of numbers of CSN retrogradely labeled from C1 in Crim1 WT, heterozygous, and null mice. There is no difference between groups.

Scale bars: 500 μm.

In this study, we functionally investigate two top candidate controls that were identified using this differential transcriptomics approach: Crim1 and Kelch-like 14 (Klhl14), genes for which there are no previously known functions in the central nervous system (CNS). In the accompanying study (Sahni et al., 2021), we identified that both these controls show remarkably specific expression by segmentally distinct CSN subpopulations during development; Crim1 is a transmembrane protein expressed by CSNTL, while Klhl14 belongs to the broad-complex, tramtrack, and bric a brac (BTB)-Kelch superfamily of proteins and is specifically expressed by CSNBC-lat. Both genes also exhibit highly dynamic spatiotemporal expression: Crim1 expression by CSNTL peaks at P4, coincident with the peak of CSNTL axon extension to distal spinal segments, while Klhl14 expression by CSNBC-lat peaks early postnatally from E18.5 to P1, when these axons limit extension to proximal segments. We now employ gain- and loss-of-function approaches to identify that Crim1 regulates thoracolumbar, while Klhl14 regulates bulbar-cervical, CSN axon targeting, respectively. These results provide tools for molecular investigation of the development of diverse yet specific functional output connectivity of cortex. Further, these results also provide the foundation for potentially investigating molecular bases of segmental targeting of other descending, spinal-projecting pathways.

RESULTS

CSNBC-lat reside outside ‘‘classic’’ motor cortex (CSNBC-lat) and project only to BC segments. CSN in medial sensorimotor cortex project either only to BC segments (CSNBC-med) or past cervical to thoracolumbar segments (CSNTL) (Sahni et al., 2021). These CSN subpopulations continue to exhibit distinct spinal axon projection and gray matter collateralization into maturity and are molecularly distinct well before final connectivity is established. Crim1 is expressed by CSNTL, while Klhl14 is expressed by CSNBC-lat (Sahni et al., 2021; Figure 1B). This enables molecular delineation and prospective identification of these CSN subpopulations: (1) Klhl14 expression delineates Klhl14-positive CSNBC-lat from Klhl14-negative CSNBC-med; (2) all CSNTL are Klhl14 negative; and (3) ~95% of CSNTL express Crim1 (Sahni et al., 2021; schematized in Figure 1A). This early axon extension specificity between CSNBC and CSNTL suggests early molecular specification; distinct subpopulations are molecularly controlled either to extend axons past thoracic T2 or to limit axon extension to cervical cord. We investigated whether these top candidate molecular controls—Crim1 and Klhl14—regulate differential axon targeting during development.

CSN are specified normally and project normally to the spinal cord in Crim1-null mice

We first investigated Crim1 function in CSNTL axon targeting. We generated Crim1 nulls (Figures 1C–1C′′) by outbreeding Crim1GCE mice (Harding et al., 2011; STAR Methods) to a CD1 background to generate Crim1GCE homozygotes, similar to Pennisi et al. (2007). Crim1 nulls are smaller than wild-type (WT) or heterozygous littermates (Figures S1A and S1Aʹ) and survive to P28 with some perinatal lethality.

We first investigated overall CSN specification in Crim1 nulls. Given predominantly postnatal expression of Crim1, projection neuron specification occurs normally in Crim1 nulls. Subcerebral projection neuron-specific controls Fezf2 (Figures 1D–1D′′) and CTIP2 (Figures S1B and S1Bʹ) are expressed normally, as are callosal projection neuron (CPN)-specific SATB2 (Figures S1C and S1Cʹ) and corticothalamic projection neuron (CThPN)-specific FOG2 (Figures S1D and S1Dʹ). CSNmedial-specific and CSNBC-lat-specific controls are also unaltered in Crim1 nulls: CSNmedial-specific genes such as Cry-mu (Figures 1E–1E′′) and St6galnac5 (Figures S1E and S1Eʹ) are expressed normally, as are CSNBC-lat-specific Klhl14 (Figures S1F and S1Fʹ) and Cartpt (Figures S1G and S1Gʹ). This indicates that CSN specification and differentiation (including CSNTL in particular) occur normally without Crim1 function. Consistent with normal overall molecular differentiation of CSN, retrograde labeling from cervical cord reveals that CSN axon extension to cervical C1 is unaffected without Crim1 function (Figures 1F–1J).

Crim1 is required for CSNTL axon extension toward thoracolumbar spinal segments

We next investigated developmental axon extension in Crim1 nulls by what normally would have been CSNTL. Retrograde labeling from lumbar/caudal thoracic cord (T13-L1) at P5 (when CSN axons first reach this level) revealed significantly fewer labeled lumbar-projecting CSN (CSNL) in Crim1 nulls (Figures 2A–2Iʹ and 2J). We investigated whether other CSN might ‘‘compensate’’ for this deficit but found that there are no retrogradely labeled neurons outside the normal location of CSNTL in medial cortex. Taken together, these results indicate that while CSN are normally specified and express other CSNmedial genes normally in Crim1 nulls, they fail to extend axons to caudal thoracolumbar segments from the earliest developmental stages.

Figure 2.

Crim1 functions to direct CSNTL axons specifically toward caudal thoracolumbar segments

(A–I) Coronal sections (from rostral to caudal) of P7 brains retrogradely labeled from T13/L1 at P5. (Aʹ– Iʹ) Magnified view of regions boxed in (A)–(I). There is a striking reduction in number of retrogradely labeled CSN (subset of CSNTL; labeled ‘‘CSNL’’) in Crim1 nulls.

(J) Quantification of number of retrogradely labeled CSNL in Crim1 WT, heterozygous, and nulls. There are significantly fewer retrogradely labeled CSNL in Crim1 nulls at both P5 and P28 (at P5, p = 0.0028 by one-way ANOVA; Crim1 nulls are significantly different from Crim1 WT [p < 0.05 by Tukey’s HSD test] and Crim1 heterozygotes [p < 0.01 by Tukey’s HSD test]; at P28, p = 0.0035 by one-way ANOVA; Crim1 nulls are significantly different from Crim1 WT [p < 0.01 by Tukey’s HSD test] and Crim1 heterozygotes [p < 0.05 by Tukey’s HSD test]).

(K) AAV-EGFP injected at P0 into medial cortex to visualize CSNTL axon projections at P4.

(L–O) Axial views of the dorsal funiculi in Crim1 WT (L and N) and Crim1 nulls (M and O) at C1-C2 (L and M) and T1-T2 (N and O). Red dotted lines indicate dorsal funiculus. There is no visual difference between groups.

(P and Q) Composite image created by overlaying a series of horizontal sections of the thoracic cord from WT (P) and Crim1 nulls (Q). WT CSNTL axons extend through thoracic cord to its caudal-most limit, reaching lumbar cord. While a significant number of Crim1-null axons enter the thoracic cord, the majority of them fail to extend substantially caudally, and very few axons reach caudal thoracic levels. CST intensity drops substantially by only a quarter of the distance into the thoracic cord in Crim1 nulls (arrows).

(R) Quantification of fluorescence intensity of labeled CST axons in dorsal funiculus at T1-T2 normalized to intensity at C1-C2. There is no significant difference between groups.

(S) Quantification of fluorescence intensity of labeled CST axons in dorsal funiculus of horizontal sections of the thoracic cord at specific distances (from rostral to caudal) in WT and Crim1 nulls. CST axon intensity significantly decreases by 25% of the distance into the thoracic cord in Crim1 nulls (p = 0.0125 by two-way ANOVA with repeated measures, *p < 0.05 by Fishers LSD).

Scale bars: (A–I) 500 μm; (K) 1 mm; (L–Q) 100 μm.

We next confirmed and extended these results via anterograde labeling. We injected AAV expressing EGFP into caudomedial cortex at P0 (Figure 2K) to visualize CSNTL axons at P4. This enabled investigation of how far caudal CSNTL axons project in the cord. We find that CSNTL axons extend normally from cervical C1 to thoracic T2 in Crim1 nulls (Figures 2L–2O). Quantification of labeled corticospinal tract (CST) fluorescence intensity finds no difference at T1–T2 (normalized to intensity at C1–C2; Figure 2R) in Crim1 nulls compared to Crim1 WT; Crim1-null CSNTL axons appear to extend normally through the cervical cord to reach T2.

In striking contrast, Crim1-null CSNTL axons exhibit significant reduction in extension from T2-L1 (Figures 2P and 2Q). We find a significant (~20%) reduction in CST intensity in Crim1 nulls even by only 25% of the distance through the thoracic cord. This reduction persists until almost no axons reach the thoracolumbar junction (Figure 2S). In WT mice, ~20% (19% ± 4%) of axons that enter the thoracic cord extend to lumbar L1, and this axon population is almost completely absent in Crim1 nulls (intensity at T13-L1 in Crim1 nulls: 2% ± 1%). Therefore, these early lumbar projections are almost completely eliminated in Crim1 nulls, suggesting that Crim1 functions critically in what are likely the pioneer lumbar-projecting axons to enable their extension through thoracic and into lumbar cord. These data indicate that Crim1 is specifically required by CSNTL to extend axons through thoracic into lumbar spinal segments.

Reduction of axon extension to thoracolumbar segments in Crim1-null mice persists into maturity

We next investigated whether reduction in CSN axon extension to thoracolumbar segments at these early developmental times is maintained. We performed retrograde labeling from T13-L1 at P23 and analyzed the mice at P28. We find that early reduction in CSN axons reaching T13-L1 persists in P28 mice (Figure 2J).

Next, to rigorously investigate whether CSN remain in medial cortex, or whether they might have been eliminated potentially due to lack of targeting, we investigated expression of cardinal CSN-specific genes at P28 in Crim1 nulls. We find that normal molecular development of CSN is maintained; subcerebral-specific Fezf2 (Figures S2A–S2Bʹ) and CSNmedial-specific Cry-mu (Figures S2C–S2Dʹ) are expressed normally in P28 Crim1-null cortex. Therefore, even though CSN persist in more mature cortex in the position normally held by CSNTL, they fail to extend axons to caudal thoracolumbar segments, even at maturity. This indicates that Crim1 enables CSNTL axon extension from early development at P4 into maturity at P28.

Crim1 functions cell autonomously in CSNTL to enable axon extension to thoracolumbar spinal segments

Crim1 is expressed by developing spinal motor neurons (Kolle et al., 2000) and, we find, also by other spinal neurons (Sahni et al., 2021). CSNTL axon extension deficits in Crim1 nulls might therefore arise from CSN-intrinsic and/or CSN-extrinsic Crim1 function. Arguing for centrally CSN-intrinsic function, the reduction in CSNTL axon extension at P4 occurs prior to CSN axons encountering spinal neurons, while they are still traversing the dorsal funiculus. This suggests that axon extension deficits in Crim1 nulls are not centrally due to spinal Crim1 expression. We directly investigated the possibility of additional, likely smaller, non-CSN-intrinsic effects of spinally derived Crim1 by evaluating potential rescue by cortical/CSN Crim1 overexpression in Crim1 nulls.

We used AAV-mediated gene delivery to express either EGFP alone (control) or EGFP plus Crim1 (potential rescue). As controls, we injected AAV particles engineered to express EGFP into medial cortex at P0 in both WT (‘‘WT’’ in Figure 3A) and Crim1 nulls (‘‘KO’’ [knockout] in Figure 3Aʹ). In other Crim1 nulls, we co-injected AAV particles engineered to express Crim1 along with AAV EGFP (‘‘KO + Crim1’’ in Figure 3A′′). In separate experiments, we confirmed that this enables cortical Crim1 overexpression (AAV-Crim1; Figures 4J and 4K). All mice were perfused at P7, and we investigated CSNTL axon extension in the dorsal funiculus at C1-C2, T1-T2, and L1-L2 in WT (Figures 3B–3D), KO (Figures 3E–3G), and KO + Crim1 (Figures 3H–3J) mice. Quantification at T1-T2 reveals no difference among all three groups, indicating normal CSN axon extension to T1-T2 in Crim1 nulls (Figures 3C, 3F, and 3I). This confirms, at P7, our earlier findings at P4 (Figure 2R) that Crim1 is not required for CSNTL axon extension to T2.

Figure 3.

Crim1 functions predominantly cell autonomously in CSNTL to enable axon extension to thoracolumbar segments

(A–A′′) Whole-mount images of P7 brains from WT (A) and Crim1 nulls (KO) (Aʹ) injected in medial cortex at P0 with AAV engineered to express EGFP (AAV-EGFP). (A′′) Whole-mount image of a P7 Crim1-null brain co-injected at P0 with AAV2/1 engineered to overexpress Crim1 (AAV-Crim1) along with AAV-EGFP (KO + Crim1).

(B–Jʹ) Axial spinal sections from the same P7 mice at cervical (B, E, and H), thoracic (C, F, and I), and lumbar (D, G, and J) segments. There is no difference in CSN axon extension from C1-C2 to T1-T2 among all three groups. In striking contrast, there is a substantial reduction in the number of CSN axons that reach L1-L2 in KO mice compared to WT controls (D and Dʹ versus G and Gʹ) or KO + Crim1 mice (J and Jʹ). This indicates ‘‘rescue’’ by cortical expression of Crim1.

(K) Quantification of fluorescence intensity of labeled CST axons in dorsal funiculus at L1-L2 normalized to intensity at C1-C2. KO mice have very substantially (>50%) reduced CST intensity compared to WT mice (p = 0.005 by Student’s t test). There is no difference in CST intensity between WT and KO + Crim1 (‘‘rescue’’) mice.

Scale bars: (A–A′′) 1 mm; (B–J) 100 μm.

Figure 4.

Crim1 is sufficient to redirect CSNBC-lat axons past cervical cord to caudal spinal segments

In (A)–(I), plasmids, designed to overexpress EGFP alone (control-EGFP) or Crim1 and EGFP (Crim1-EGFP), were delivered to developing CSNBC-lat in lateral cortex using in utero electroporation at E12.5, with tissue collected at P4 for analysis.

(A–B′′′) Assembled composite images created by overlaying a series of sagittal thoracic cord sections from control-EGFP (A) and Crim1-EGFP (B) electroporated mice. The few CSNBC-lat axons entering thoracic cord in control mice terminate within the first few thoracic segments and do not reach farther than 25% of the distance into thoracic cord (Aʹ and A′′). Crim1-EGFP-expressing CSNBC-lat axons extend significantly farther (arrows in Bʹ and B′′) with a few individual axons extending as far as 75% of the distance into thoracic cord (arrows in B′′′).

(C) Quantification of fluorescence intensity of EGFP+ axons in either control EGFP or Crim1-EGFP electroporated mice at four sites along the rostro-caudal extent of the thoracic cord. Each data point represents average normalized intensity ± SEM (n = 3 mice per group; p = 0.0002 by two-way ANOVA with repeated measures, *p < 0.001 by Fishers LSD test).

(D–Iʹ) Coronal sections of a single P4 brain electroporated in utero at E12.5 with Crim1-EGFP. (D, F, and H) EGFP fluorescence (green) shows electroporated area in lateral cortex. (E) In situ hybridization for Crim1 cDNA showing Crim1 overexpression on the electroporated side (magnified in Eʹ). Crim1 is normally restricted to medial cortex, and there is no upregulation of endogenous Crim1 in Crim1-overexpressing CSNBC-lat (G, magnified in Gʹ). Also, Klhl14 expression in Crim1-overexpressing CSNBC-lat remains unchanged (I, Iʹ).

(J–M′′′) AAV-Crim1 co-injected with AAV-EGFP (reporter) into rostrolateral cortex at P0. AAV-EGFP alone is the control. Tissue was collected at P14. (J–Kʹ) Coronal section of a brain injected with AAV-EGFP+AAV-Crim1 showing injection site in rostrolateral cortex (J). In situ hybridization for Crim1 cDNA shows overexpressed Crim1 in injection site (K, magnified in Kʹ). (L–M′′′) Assembled composite images created by overlaying a series of thoracic cord horizontal sections from control (L) and AAV-EGFP + AAV-Crim1 (M) (Crim1O/E) injected mice. Labeled axons in Crim1 O/E mice extend substantial distances into thoracic cord (magnified images in Mʹ–M′′′), with few axons reaching the caudal-most levels of thoracic cord (arrows in M′′ and M′′′).

(N and O) Inverted, high-contrast monochrome tracings from (L) and (M) of EGFP+ axons. Blue outlines the cord, and black pixels show EGFP+ axons (details in STAR Methods).

Scale bars: (A, B, L, and M) 100 μm; (D–K) 500 μm.

We next investigated whether cortical-derived Crim1 can rescue CSN axon extension to distal segments in the thoracolumbar cord (T13-L1) in Crim1 nulls. At lumbar L1, there is clear qualitative reduction in the number of CSN axons present in Crim1 nulls compared to WT (Figures 3D and 3G). This confirms the findings at P4 (Figures 2P, 2Q, and 2S) that Crim1 is necessary for CSNTL axons to extend through thoracic cord to lumbar segments. Strikingly and importantly, this reduction is not observed in KO + Crim1 mice (Figure 3J). Quantification of CST fluorescence intensity at L1-L2 reinforces these qualitative results. We find significant reduction in CSN axon extension to L1-L2 in Crim1 nulls that received AAV-GFP alone (KO) compared to WT mice (CST intensity at L1-L2 normalized to C1-C2 in WT = 28% ± 2%, KO = 13% ± 2%; Figure 3K). In contrast, CST intensity at L1-L2 in KO + Crim1 mice is statistically indistinguishable from WT mice (CST intensity at L1-L2 in KO + Crim1 = 21% ± 7%; Figure 3K). This indicates that Crim1 can CSN-autonomously rescue axon extension deficits to caudal thoracolumbar segments in Crim1 nulls. The rescue might be even closer to WT than reported here, because while one of four mice in the KO + Crim1 group appears to be an obvious outlier, the distribution of normalized CST intensity at L1-L2 in the other three KO + Crim1 mice lies entirely within the WT range (Figure 3K). In the single outlier, the lumbar CST intensity is dramatically lower than any other mouse in any experimental or control Crim1-null subgroup. Though we include this outlier in the results above, it dramatically lowers the overall average CST intensity at L1-L2 in KO + Crim1 mice (if this one outlier were excluded, the average CST intensity at L1-L2 in KO + Crim1 mice would be 28% ± 3%, almost identical to average WT intensity, 28% ± 2%). Together, the timing of effect on CSNTL axon extension combined with the rescue experiments reinforce that Crim1 functions centrally intrinsically within CSN to enable CSNTL axon extension to thoracolumbar segments.

Crim1 misexpression in CSNBC-lat redirects their axons to caudal spinal segments

We next investigated whether Crim1 can redirect CSNBC-lat axons to thoracolumbar segments. We misexpressed Crim1 in developing CSNBC-lat at two distinct times and investigated their axon extension. We first introduced a Crim1 expression construct into developing CSNBC-lat via in utero electroporation at E12.5. While Crim1 misexpression does not increase the number of CSNBC-lat axons simply entering thoracic cord at T1-T2, Crim1 misexpression in CSNBC-lat does redirect the majority of these axons past their normal rostral targets to distal thoracic segments (Figures 4A–4C).

Crim1-overexpressing CSNBC-lat (Figures 4D and 4E) do not activate endogenous Crim1 expression (Figures 4F and 4G; in situ probes for detecting endogenous versus overexpressed Crim1 are in Table S1). Further, in utero Crim1 overexpression does not alter Klhl14 expression (Figures 4H and 4I), suggesting that CSNBC-lat differentiation is not affected by Crim1 overexpression. However, there still remained the unlikely possibility that the effect on CSNBC-lat axon extension is due to alterations in early CSNBC-lat specification other than Klhl14, since overexpression by in utero electroporation begins in progenitors and continues into postmitotic neurons. We therefore next performed Crim1 overexpression in CSNBC-lat at P0 to directly investigate whether CSNBC-lat axons can be redirected post-mitotically to thoracolumbar segments once their axons have reached the cord. We used AAV-mediated gene delivery to misexpress Crim1 in CSNBC-lat. We injected AAV-Crim1 plus AAV-EGFP into rostrolateral cortex at P0 (Figures 4J and 4K). Crim1 overexpression by CSNBC-lat is sufficient to direct many of their axons to thoracolumbar segments. While few axons from control (AAV-EGFP alone) CSNBC-lat enter the thoracic cord, and no axons extend even 25% through the thoracic cord (Figures 4L, 4N, and 4U; n = 3 mice), axons of AAV-Crim1-overexpressing CSNBC-lat enter (Figures 4M, 4Mʹ, and 4O) and often traverse from 50% to the entire length of the thoracic cord (Figures 4M and 4O; n = 3/3 Crim1-overexpressing mice display CSNBC-lat axons halfway through thoracic, Figure 4M′′; and two of them display axons through the entire length of the thoracic cord, Figure 4M′′′). Together, these experiments—Crim1 loss-of-function, rescue of CSNTL axon extension by Crim1 overexpression in Crim1-null cortex, and Crim1 misexpression redirecting CSNBC-lat axons—indicate that Crim1 centrally, and predominantly cell intrinsically, controls CSN axon extension to caudal thoracolumbar segments. These data also reveal that Crim1 does not mediate CSN axon extension to thoracic T2. Rather, Crim1 regulates CSN axon extension past T2 into thoracolumbar cord.

Reduction in Klhl14 function causes aberrant CSNBC-lat axon targeting beyond the cervical cord

We investigated Klhl14 function in CSNBC-lat axon targeting as schematized in Figure 5A. We first introduced a short hairpin RNA (shRNA) construct that effects ~85% knockdown of Klhl14 expression into lateral cortex using in utero electroporation to disrupt Klhl14 expression in developing CSNBC-lat (Figures 5A, S3A, and S3B). We first confirmed that EGFP+ neurons in both control and Klhl14-shRNA-electroporated mice were located laterally throughout the rostro-caudal extent of sensorimotor cortex (Figures S3D and S3E); quantification established no fluorescence in medial cortex in either group (Figure S3F; i.e., electroporations specifically targeted CSNBC-lat, enabling investigation of CSNBC-lat axon targeting). Klhl14 shRNA does not appear to disrupt neuronal differentiation or lamination (e.g., Klhl14-shRNA-expressing CSNBC-lat express normal levels of CTIP2; Figure S3C). CSNBC-lat axon collateralization in the P4 cervical cord (C2-T1) is similarly unaltered between control and Klhl14 shRNA-electroporated mice (data not shown).

Figure 5.

Reduction in Klhl14 function causes aberrant targeting of CSNBC-lat axons beyond the cervical cord

(A) Plasmids designed to express either control scrambled shRNA or a shRNA to knock down Klhl14 expression. shRNA sequences were placed 3′ to an EGFP coding sequence. The constructs were electroporated into lateral cortex at E12.5, and tissue is collected at P4 and P14 for analysis (electroporated brains in Figure S3). Images shown are immunocytochemistry for EGFP on axial (B–E) or sagittal (F and G) sections of the cord.

(B–E) Axial sections at cervical C1-C2 (B and D) and thoracic T1-T2 (C and E) from mice that were electroporated with scrambled (B and C) or Klhl14 shRNA (D and E). (Bʹ–Eʹ) High-magnification single-plane confocal images of the dorsal funiculus (areas boxed in B–E).

(F and G) Assembled composite images created by overlaying a series of sagittal sections of thoracic cord from scrambled (F) and Klhl14 shRNA (G) electroporated mice. Klhl14-shRNA-expressing CSNBC-lat axons extend significantly farther into thoracic cord (arrows in G).

(H–Iʹ) Magnified views of boxed regions in (G) showing Klhl14-shRNA-expressing CSNBC-lat axons aberrantly extending collaterals in thoracic cord (arrowheads Hʹ and Iʹ).

(J) Quantification of percentage of axons in the dorsal funiculus at C1-C2 that extend to T1-T2. Graphs represent average percentages ± SEM (n = 4 mice for each group [control and Klhl14 shRNA]; *p < 0.05 by Student’s t test).

(K) Quantification of normalized fluorescence intensity of labeled EGFP+ axons in the dorsal funiculus on all sagittal sections of thoracic cord from either control scrambled-shRNA- or Klhl14-shRNA-electroporated mice. Each data point represents the average normalized intensity ± SEM (p < 0.0001 by two-way ANOVA with repeated measures; *p < 0.001 by Fishers LSD test).

Scale bars: 100 μm.

We next investigated whether Klhl14 functions to normally restrict CSNBC-lat axon targeting in the dorsal funiculus. At P4, while almost all axons from control, scrambled shRNA-expressing CSNBC-lat extend only within cervical cord (only 6% ± 3% of CSNBC-lat axons at cervical C1 extend to thoracic T1-T2; Figure 5J), significantly more axons from Klhl14-shRNA-expressing CSNBC-lat extend to T1-T2 (21% ± 5%; Figures 5B–5E and 5J). The rare control axons that enter the thoracic cord are limited to the first few thoracic segments (assessed from T2-L1). Further, these rare control CSNBC-lat axons never extend collaterals in the thoracic cord (Figure 5F).

In striking contrast, many axons from Klhl14-shRNA-expressing CSNBC-lat extend significantly further, reaching as far as 75% of the distance through thoracic cord (Figures 5G–5Iʹ and 5K), and they extend branching collaterals even at mid-thoracic levels (Figures 5Hʹ and 5Iʹ).

These aberrantly extended CSNBC-lat axons remain in thoracic cord even at P14 (Figures S4K–S4Nʹ). Because this is past the period of normal Klhl14 expression in CSNBC-lat, this indicates that early Klhl14 control over CSNBC-lat axon extension is critical in establishing CSNBC-lat-specific connectivity in the mature CNS. Consistent with CSN specificity (in particular, CSNBC-lat) of Klhl14 expression, there are no axon targeting defects by CPN (Figures S4E–S4H) in Klhl14-shRNA-electroporated cortex. Taken together, these results identify that Klhl14 specifically limits CSNBC-lat axon extension and branching proximal to T2.

Klhl14 functions in postmitotic CSNBC-lat specifically when CSN axons are extending into the spinal cord

Because CSNBC-lat and CSNmedial do not exhibit different rates of axon extension (Sahni et al., 2021), it is not likely that Klhl14 limits CSNBC-lat axon extension to the cervical cord by inhibiting the rate of axon extension. We directly investigated this unlikely, but still theoretical, possibility. We examined control-versus Klhl14-shRNA-expressing CSNBC-lat axons at earlier developmental times in case a reduction of Klhl14 function causes CSNBC-lat axons to grow faster into thoracic cord, thus past their normal targets. At P0, when CSN axons first reach the cord, there is no difference in length of axon growth betweencontrol-shRNA- and Klhl14-shRNA-expressing CSNBC-lat (Figures S4A–S4D). These data indicate that Klhl14 does not regulate rate of axon extension. Rather, these data indicate that Klhl14 regulates the specificity of CSNBC-lat axon targeting (i.e., to restrict CSNBC-lat axons from extending beyond T2).

These results also suggest that Klhl14 functions while CSNBC-lat axons are extending in the cord, consistent with the time course of Klhl14 expression in CSNBC-lat: Klhl14 expression peaks early at E18.5 and P1, then gradually declines from P4 to P10, and expression is completely absent by P14 (Sahni et al., 2021). However, there remained the unlikely possibility that Klhl14 shRNA alters CSNBC-lat differentiation before their axons reach the cord, ultimately causing aberrant axon extension later at P4. To directly investigate this unlikely possibility, and to delineate the specific temporal requirement of Klhl14 function, we generated conditional Klhl14 mice in which the second coding exon of Klhl14 is flanked by loxP sites.

To conditionally ablate Klhl14 function postmitotically, we co-injected AAV-Cre and AAV-FLEX-tdTomato at P0 into lateral cortex of Klhl14flox/flox (Klhl14 conditional KO [cKO]) and Klhl14 WT as controls (Figure 6A). This manipulation leaves Klhl14 function intact through birth and early postmitotic differentiation of CSNBC-lat; Klhl14 function is ablated only as CSNBC-lat axons are reaching the cord. We investigated CSNBC-lat differentiation and axon extension at P7 using the AAV-FLEX-tdTomato reporter (tdTomato is expressed only by neurons also receiving AAV-Cre). We first confirmed deletion of Klhl14 expression in lateral cortex in AAV-Cre-injected hemisphere in Klhl14 cKO mice (Figures S5A and S5B). Fully consistent with results using Klhl14 shRNA, this postmitotic conditional deletion of Klhl14 does not alter CTIP2 expression and, thus, does not disrupt overall subcerebral projection neuron differentiation of CSNBC-lat (Figure S5C).

Figure 6.

Klhl14 functions in postmitotic CSNBC-lat early postnatally, specifically when CSN axons are extending into cervical cord

(A) AAV-Cre and AAV-FLEX-tdTomato were co-injected into rostrolateral cortex at P0 in Klhl14 WT control or Klhl14flox/flox mice to conditionally ablate Klhl14 function in CSNBC-lat (Klhl14 cKO). Mice were perfused at P7, and axons were visualized using immunocytochemistry for tdTomato.

(B and C) Whole-mount images of P7 brains from Klhl14 WT and Klhl14 cKO mice that were injected with AAV-Cre + AAV-FLEX-tdTomato at P0. tdTomato fluorescence (red) demarcates closely matching injection site and extent in all mice (quantification in Figure S5).

(D–Gʹ). Axial sections at cervical C1-C2 (D and F) and thoracic T1-T2 (E and G) from either WT (D and E) or Klhl14 cKO mice (F and G). (Dʹ–Gʹ) High-magnification single-plane confocal images of the dorsal funiculus (areas boxed in D–G). Substantially more axons extend to T1-T2 in Klhl14 cKO mice compared to controls.

(H) Quantification of fluorescence intensity of tdTomato+ CSNBC-lat axons at T1-T2 normalized to C1-C2 in Klhl14 WT versus Klhl14 cKO mice. Graphs show average percentages ± SEM (n = 4 mice for each group [WT and Klhl14 cKO]; *p < 0.05 by Student’s t test).

(I–J′′) Assembled composite images created by overlaying a series of horizontal sections of thoracic cord from either Klhl14 WT (I) or Klhl14 cKO (J) mice showing strikingly extended, aberrant CSNBC-lat projections in thoracic cord. (Jʹ and J′′) Monochrome, magnified views of boxed region in (J) at two separate Z levels, showing aberrant gray matter collaterals at these caudal thoracic levels in Klhl14 cKO mice.

(K–L′′) Higher magnification views of boxed regions in (Jʹ) and (J′′) showing these collaterals (arrowheads K and L–L′′).

Scale bars: (B and C) 1 mm; (D–G, I, and J) 100 μm.

We next investigated CSNBC-lat axon extension in the cord. We first confirmed the consistency of AAV injections in lateral cortex in WT and Klhl14 cKO mice via whole-mount images (Figure 6B and 6C). We verified that there was no labeling in medial cortex in WT and Klhl14 cKO mice (Figures S5D–S5F). We next investigated CSNBC-lat axon extension in WT and Klhl14 cKO mice by analyzing tdTomato+ axons in dorsal funiculus at cervical C1-C2 and thoracic T1-T2. At P7, consistent with our prior anterograde analyses, >95% of all CSNBC-lat axons in WT mice terminate in the cervical cord, with only rare axons present at T1-T2 (Figure 6D and 6E). In contrast, significantly more CSNBC-lat axons project to T1-T2 in Klhl14 cKO mice (Figures 7F and 7G). Quantification of CST fluorescence intensity at T1-T2 reveals a 2.8-fold increase in the number of CSNBC-lat axons that extend to T1-T2 in Klhl14 cKO versus WT mice (CST intensity at T1-T2 normalized to C1-C2: Klhl14 WT = 3.2% ± 1.2%; Klhl14 cKO = 9.3% ± 1.5%; Figure 6H). This indicates that significantly more CSNBC-lat axons extend into thoracic cord upon postmitotic conditional deletion of Klhl14. These aberrantly extended CSNBC-lat axons in Klhl14 cKO mice also collateralize in the thoracic gray matter (Figure 6I–7Lʹ), similar to Klhl14-hRNA-expressing CSNBC-lat.

Figure 7.

Reduction or deletion of Klhl14 function causes ectopic Crim1 expression in lateral cortex

(A–B′′) Coronal section of a P4 brain electroporated in utero at E12.5 with Klhl14 shRNA. (A) EGFP fluorescence (green) shows electroporation in lateral cortex. (B) In situ hybridization on the same section showing Crim1 expression. Crim1 is normally restricted to medial cortex. However, Klhl14 shRNA causes ectopic Crim1 expression (arrowheads) in lateral cortex (Bʹ), only in electroporated hemisphere (compare Bʹ and contralateral B′′).

(C–D′′) Coronal section of a P7 Klhl14 cKO brain injected with AAV-Cre + AAV-FLEX-tdTomato at P0. (C) TdTomato fluorescence (red) shows injection in lateral cortex. (D) In situ hybridization on the same section showing Crim1 expression. Similar to Klhl14 shRNA, postmitotic Cre-dependent Klhl14 deletion causes ectopic Crim1 expression (arrowheads) in lateral cortex (Dʹ), only in injected hemisphere (compare Dʹ and contralateral D′′.

Scale bars: 500 μm.

The shRNA and cKO results both indicate very similar increases in CSNBC-lat axon extension to thoracolumbar segments using two independent approaches (~3.5-fold using shRNA, ~2.8-fold using the conditional null). Further, the results with Cre-dependent conditional gene deletion in Klhl14 cKO mice delineate a temporal, postmitotic requirement of Klh14 function in regulating CSNBC-lat axon extension, when these decisions are being executed in the spinal cord.

Reduction or deletion of Klhl14 function causes ectopic Crim1 expression in lateral cortex

We also investigated whether this aberrant CSNBC-lat axon extension into thoracic cord with loss of Klhl14 function might be accompanied by other CSNTL-like changes (e.g., changes in Crim1 expression). We find that Crim1 is aberrantly expressed in lateral cortex at P4 in Klhl14-shRNA-electroporated mice (Figure 7A and 7B; compare 7Bʹ with 7B′′). However, not all CSNTL genes are aberrantly expressed (e.g., CSNmedial genes Cry-mu and Cdh8 remain restricted to medial layer V after Klhl14-shRNA electroporation; Figures S6A–S6D). Klhl14 cKO mice display similar, aberrant Crim1 expression in lateral cortex in the AAV-Cre-injected hemisphere (Figure 7C and 7D; compare 7Dʹ with 7D′′). These results by two independent experimental approaches to Klhl14 loss-of-function indicate that the population of CSNBC-lat becomes more CSNTL-like upon loss of Klhl14 function, in both its aberrant axon extension and gene expression.

Finally, to investigate whether Klhl14 might be sufficient to restrict axon targeting to cervical cord, we misexpressed Klhl14 in medial cortex. We injected AAV-Klhl14 into medial cortex and examined CSNTL axon extension to thoracolumbar targets. We find no effect on CSNTL axon extension as assessed by both anterograde (Figures S7C–S7E) and retrograde (Figures S7A and S7B) analyses. These data additionally indicate that Klhl14 does not control CSNBC-lat axon elongation by simply regulating rate of axon extension. Together, these results indicate that Klhl14 is necessary but not sufficient to limit CSNBC-lat axons to cervical cord.

DISCUSSION

Segmentally precise CSN connectivity enables highly skilled motor control (and related ‘‘top-down’’ sensorimotor feedback and autonomic functions). Further, CSN are clinically relevant; individual human neurodegenerative diseases predominantly affect segmentally specific CSN (e.g., bulbar amyotrophic lateral sclerosis [ALS] or hereditary spastic paraplegias [HSPs]). The molecular bases for development of such segmental specificity by CSN are not known.

In this report and the accompanying paper (Sahni et al., 2021), we identify that distinct CSN subpopulations are molecularly specified during development to target distinct spinal segments at maturity. We identify molecular controls that distinguish during development anatomically, hodologically, and likely later functionally distinct CSN subpopulations, and they govern differential axon targeting of CSN at T2. These controls establish the first stages of a dual-directional set of complementary controls over critical motor circuitry.

Mechanisms controlling differential axon extension by CSN subpopulations

Crim1 and Klhl14 have no previously reported function in the CNS. Gain- and loss-of-function results indicate that Crim1 does not mediate CSN axon extension to T2 but affects axon extension past T2 into thoracic cord. Axon extension abnormalities are apparent by ~25% of the rostro-caudal distance through thoracic cord and are maintained more distally, with a small subset of CSNTL axons extending to lumbar cord (Figure 2). Crim1 might function in ‘‘pioneer’’ lumbar-projecting CSN axons to enable extension through thoracic into lumbar cord. However, Crim1 appears to not function alone; retrograde labeling at P28 identified only ~50% reduction in CSN projections to lumbar cord in Crim1 nulls (Figure 2J), indicating that some CSNTL axons extend to distal targets independent of Crim1. Interestingly, spared lumbar-projecting CSN without Crim1 function reside in the typical medial location, indicating that there is no anatomically distinct, alternate Crim1-independent lumbar-projecting CSN subpopulation. Crim1 rescue in Crim1 nulls indicate that Crim1 functions autonomously in CSN.

Crim1 is a transmembrane protein with a large extracellular domain containing cysteine-rich repeats (CRRs) that can bind multiple growth factors (e.g., bone morphogenetic proteins [BMPs] and transforming growth factor β [TGF-β]). BMPs can function in axon guidance (Charron and Tessier-Lavigne, 2005; Salie et al., 2005). The Crim1 ortholog in Drosophila motor neurons controls growth of neuromuscular junctions, regulating BMP levels (James and Broihier, 2011). Crim1 might enable or direct CSNTL axon extension to distal segments by acting as a guidance receptor. Crim1 has a predicted IGF-binding motif (Kolle et al., 2000) that might augment IGF-I signaling to drive CSNTL axon extension (Ozdinler and Macklis, 2006). Crim1 levels in CSNTL peak during axon extension to distal targets (Sahni et al., 2021) and are sufficient to direct thoracolumbar axon extension by CSNBC-lat (Figure 4). Both expression and temporal control of Crim1 levels might serve to direct long CSNTL axon extension. Alternatively, Crim1 might cause CSNTL axons to not recognize the cervical cord as an appropriate target region and, thus, to extend to distal targets, potentially by recognition of cues in the extracellular matrix. Crim1 is known to regulate b1 integrin-mediated cell adhesion in the developing lens (Zhang et al., 2016); a similar mechanism might regulate CSNTL axons.

We identify that Klhl14 limits CSNBC-lat axons to cervical cord. Klhl14 might function in multiple distinct ways: by instructing CSNBC-lat axons to not cross T2; by instructing CSNBC-lat axons to recognize brainstem and cervical cord as appropriate targets; or a combination of these mechanisms. Though Kelch family members are reported to regulate cytoskeletal dynamics, this does not appear to be a mechanism of Klhl14 limiting CSNBC-lat axon extension.

Crim1 expression during early development largely predicts thoracolumbar projection at maturity (CSNTL) (Sahni et al., 2021). Crim1 might be transcriptionally repressed in CSNBC. Consistent with this, Crim1 overexpression redirects a subset of CSNBC-lat axons to caudal thoracic cord. Loss of Klhl14 function by two distinct approaches results in aberrant CSNBC-lat axon extension into thoracic cord; this is accompanied by ectopic Crim1 expression in lateral cortex. This suggests that Crim1 regulation might be critical to limit CSNBC-lat axons to cervical cord.

In both Klhl14 loss-of-function and Crim1 gain-of-function experiments, even those CSNBC-lat axons that aberrantly extend into thoracic cord do not exit into lumbar cord. In contrast, cortical/CSN Crim1 overexpression in Crim1 nulls is sufficient to direct CSNTL axons to caudal thoracic segments and even to extend past thoracic segments into lumbar cord. Together, these results suggest that a subset of CSNTL possesses additional mechanisms enabling them to project past thoracic into lumbar segments, if Crim1 is present to enable them to traverse the thoracic cord. This is consistent with our earlier results that lumbar-projecting CSN (CSNL) are a distinct subset within CSNTL, both spatially (most caudal among CSNTL) and, seemingly, molecularly (a subset of CSNTL genes expressed only in the caudal CSNTL domain) (Sahni et al., 2021). CSNL might rely on a subset of CSNTL molecular controls that are CSNL specific to direct or enable CSNL axons to enter the lumbar cord.

Finally, these results highlight that understanding development of segmental specificity of corticospinal circuitry benefits substantially from investigation at early developmental times, well before CSN axon collaterals invade the spinal gray matter or synaptic connectivity is even established. This is further highlighted by the fact that some molecular regulators (e.g., Klhl14) are no longer expressed past the first postnatal week, after this initial development is complete. Therefore, previous or future work not focused on these early developmental stages would be predicted to not identify at least some important early regulators of CSN axon targeting. This is especially relevant to recent work investigating neuronal diversity, particularly in adult motor cortex via single-cell transcriptomics (Yao et al., 2020), since these approaches after early development has ended would be predicted to not identify such transiently expressed early regulators, even if sequencing depth were sufficient.

Differential CSN axon targeting at thoracic T2 might reflect differential CSN-extrinsic cues

The divergent axon guidance decisions by CSN subpopulations at T2 likely reflect distinct, differential responses of their growth cones to distinct environmental cues (Raper and Mason, 2010) presented at the transition between these spinal segments. The specification of columnar identity along the rostro-caudal axis in the spinal cord is mediated by Hox expression patterns in spinal motor neurons, directing differentiation of distinct motor neuron pools at distinct spinal levels, particularly limb innervating at cervical versus autonomic neurons at thoracic levels (Dasen et al., 2003). Guidance cues, potentially downstream of Hox genes, might be differentially expressed between cervical and thoracic segments. CSN axon guidance decisions occur prior to collateral extension into gray matter. Therefore, it appears likely that such cues might be presented by spinal white matter glia (e.g., molecularly distinct astrocytes along the dorsoventral axis in the cord encode positional cues to direct development of specific spinal neurons; Molofsky et al., 2014).

Implications for CSN function, plasticity, evolution, and disease vulnerability

Although the CST in rodents has been mostly thought to control forelimb movement, it also controls elements of hindlimb control (Serradj et al., 2014). Motor analysis of Crim1 nulls could be used to address some of these questions. Future investigations using Crim1 cKO mice (Chiu et al., 2012) could be employed to investigate functional consequences of a lack of CSNTL innervation in distal spinal segments.

It is intriguing to speculate whether molecular controls over segmentally specific connectivity during development might be altered and/or re-activated in instances of corticospinal plasticity in the adult CNS following disease or injury. For instance, CSN residing in hindlimb motor cortex sprout new collaterals in the cervical cord after a thoracic spinal cord injury (Ghosh et al., 2010) or after an ischemic stroke in forelimb motor cortex (Starkey et al., 2012). The results presented here might also shed light on such plasticity. While CSNTL fail to extend axons to their distal spinal targets in Crim1 nulls, they still persist into maturity. It is intriguing to speculate that these CSNTL might display aberrant connectivity in the cervical cord, potentially mimicking the effects noted after adult CST injury described above. Future investigations into spinal connectivity of misrouted CSN axons after Crim1/Klhl14 loss-of-function could begin to address such questions.

Loss of Klhl14 function results in ectopic Crim1 expression by CSNBC-lat (Figure 7), indicating what appears to be a hierarchical order of controls over CSN segmental targeting. These results raise the intriguing possibility that CSNBC-specific controls such as Klhl14 evolved to suppress potentially older, ‘‘default’’ controls such as Crim1, enabling increased numbers and diversity of CSN projecting to the most dexterous control circuits in the cervical cord and functionally distinguishing CSNBC from evolutionarily older CSNTL subpopulations. This might underlie the extraordinary evolutionary expansion of skilled forelimb movements, since even subtle changes in nervous system organization can cause large behavioral changes (Katz and Harris-Warrick, 1999).

CSN degeneration in ALS, along with degeneration of spinal motor neurons, causes spasticity and paralysis (Bruijn et al., 2004). ALS and related motor neuron disorders involving CSN do not affect all CSN equally; in bulbar forms of ALS, brainstem-projecting CSN degenerate preferentially, while in HSP, primarily lumbar-projecting CSN degenerate (Salinas et al., 2008). It appears increasingly likely that such selective vulnerability might arise, at least in part, from dysregulation of developmental control over differential axon targeting by CSN subpopulations. Consistent with this, human Crim1 maps close to a spastic paraplegia locus (Kolle et al., 2000), suggesting one such potential link to subtype-specific CSN disease. The Crim1 function identified here adds credence to the hypothesis that dysregulation of early development might underlie selective CSN vulnerability in some motor neuron disease.

The work presented in this paper is a first step toward understanding how segmentally specific corticospinal organization is initially established during development; this segmentally specific circuitry would eventually include motor, sensory, and autonomic circuit organization and control. Even if regulators of early axon targeting (e.g., Crim1, Klhl14) do not establish functionally specific (e.g., motor versus autonomic) circuits, there are likely such regulators that work combinatorially with these early controls. Therefore, axon targeting specificity likely establishes the first stage of ultimate connectivity and functional specialization.

Delineation of gene sets highly specifically expressed by CSNBC versus CSNTL reveals a molecular network controlling connectivity rostral and caudal to T2. Crim1 and Klhl14, on their own, are insufficient to fully explain this specificity. We present these as exemplar regulators of what are likely to be downstream growth-cone-located subcellular mechanisms of segmentally distinct axon targeting. Future investigations building from this work will likely identify increasingly detailed mechanisms (both CSN-intrinsic and CSN-extrinsic cues in the cord) that effect precision of axon targeting. Other identified subpopulation-specific genes have promise to elucidate additional intersectional levels of precision, potentially including additional segmental delineation, axon branching and connectivity within segments, and potentially even mono-versus bi-synaptic corticospinal-spinal connectivity. In this regard, we have functionally investigated additional candidates. These include the extracellular matrix protein Lumican as controlling spinal-segment-specific axon collateralization, rather than controlling axon targeting in spinal white matter (Itoh et al., 2021). Future investigation of interactions and hierarchical organization of these molecular regulators offers potential elucidation of development, organization, disease, and regeneration of corticospinal connectivity.

STAR★METHODS

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Dr. Jeffrey D. Macklis (jeffrey_macklis@harvard.edu).

Materials availability

We plan to deposit Klhl14flox mice to the Jackson Laboratory or MMRC. All unique/stable reagents generated in this study are available with a materials transfer agreement from the lead contact for academic, non-commercial use; negotiation and completion of a materials transfer agreement with Harvard University is required if there is potential for commercial application.

Data and code availability

Microscopy data reported in this paper will be shared by the lead contact upon request.

No original code was generated as part of this study.

Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Mice used in this study

Wild-type mice on a pure C57BL/6 background were obtained from Charles River Laboratories (Wilmington, MA). The day of vaginal plug detection was designated as E0.5. The day of birth was designated as P0. All mouse studies were approved by the Harvard University IACUC, and were performed in accordance with institutional and federal guidelines.

Crim1GCE mice (Harding et al., 2011) were obtained from Jackson Laboratories (stock number 017495); they contain an EGFP coding sequence followed by a tamoxifen-inducible Cre recombinase coding sequence placed 3′ to the ATG of the first codon. The original depositing investigator observed no EGFP expression, and we confirm no EGFP expression at any age.

To obtain Crim1GCE homozygous null mice, we backcrossed males from the original line (maintained on a C57BL/6 background) with WT CD1 females. At each generation, transgenic males were mated with WT CD1 females. Mixed background (F6) transgenic heterozygote mice were mated to generate Crim1 nulls. To distinguish Crim1GCE WT, heterozygous and homozygous null mice, the following primers were used:

WT allele

Forward – 5′- GCAGGAGGATGTACTTGGTG - 3′

Reverse – 5′ – TGTGTCTATGCGTGTTCGG - 3′

GCE allele

Forward – 5′- AAGTTCATCTGCACCACCG - 3′

Reverse – 5′ – GTTATTCGGATCATCAGCTACACC - 3′

Klhl14 conditional allele mice were generated using embryonic stem cells obtained from the KOMP Repository (https://www.komp.org). The mouse strain was created from embryonic stem cells clone EPD0735_1_H02, obtained from the KOMP Repository and generated by the Wellcome Trust Sanger Institute (WTSI). Targeting vectors used were generated by the WTSI and the Children’s Hospital Oakland Research Institute as part of the Knockout Mouse Project (3U01HG004080). Methods used on the organization and generation of CSD targeted alleles have been published (Skarnes et al., 2011). Targeted embryonic stem cells clones were injected into blastocysts at the Harvard Genome Modification Facility, and the chimeras were bred with C57BL/6J albino mice (Jackson Laboratories; stock number 00058). Positive pups with black coat color were first confirmed using genotyping PCR. Per KOMP recommendation, correct gene targeting was then confirmed in the founder and in subsequent litters from these founders using long range PCR. The initial (Klhl14-lacZ knock-in knock out) mice were then bred with R26-FlpO mice (Jackson Laboratories; stock number 12390) to obtain Klhl14 floxed mice as described (Skarnes et al., 2011), which were then bred to homozygosity. To distinguish between the Klhl14 WT versus Klhl14flox allele, the following genotyping primers were used:

Klhl14 Forward

TTCTTAGTGCCCTTTCCTCCGTACC

Klhl14 Reverse

ATGAAACTCTGGTGGCTTTGGATGC

The genders of early postnatal mice were not determined. Mice were used at the following ages:

Crim1GCE mice: were used at P4, P7, P28, and as adults.

Klhl14flox mice: were used at P7, and as adults.

METHOD DETAILS

Anterograde and retrograde labeling

Retrograde labeling was performed as described in the accompanying study (Sahni et al., 2021). Briefly, we used a pulled glass micropipet attached to a nanojector (Nanoject II, Drummond Scientific, Broomall, PA) to bilaterally inject the retrograde label Cholera Toxin B subunit (CTB; Thermo Scientific) into specific spinal levels on each side of the midline using ultrasound backscatter microscopy (Vevo 770; VisualSonics, Toronto, Canada). For cervical labeling (P3), we used the fourth ventricle in the medulla, and the spinal-medullary junction as landmarks. The landmark for labeling CSN whose axons reach thoracic T12/13 -lumbar L1 was established by examining the relative position of the vertebral column as it approaches the dorsal surface of the body (moving from caudal thoracic levels rostrally, where the column is located more deeply and closer to the viscera, to lumbar levels caudally, where the column is situated closer to the dorsal surface). The central landmarks for all intraspinal injections are as described (Sahni et al., 2021)–the midline, vertebral bodies, dorsal aspect of the spinal cord, and echo density in the dorsal funiculus. For these neonatal injections, pups were anesthetized under ice for 4 minutes. The pups were placed on a heating pad for recovery.

To retrogradely label CSN projecting to T/13-L1 in adults, we anesthetized mice using isoflurane anesthesia (2.5% isoflurane in 100% oxygen). We exposed the caudal thoracic and rostral lumbar vertebrae under standard aseptic surgical procedures (n = 3 each for Crim1GCE WT and homozygous null mice). Using L1 as a landmark, we used bone rongeurs (World Precision Instruments) to create a small window in the L1 dorsal vertebral lamina to expose the dura. A rolled gauze was placed under the mouse’s abdomen to slightly elevate the lumbar cord. The spine was gently stretched, and the tail was taped to maintain a flat spine for injection. Throughout each injection, the dura and cord were kept moist using warm, sterile saline. We injected 300 nL CTB-555 on either side of the midline using the same injection technique as for postnatal injections. After each injection, micropipettes were left in the spinal cord for an additional 1 min before withdrawl. Following each injection, wounds were closed using wound clips (AUTO-CLIP®, Becton Dickinson). Post-surgical care was performed as previously described (Greig et al., 2016). Mice were perfused 5 days later.

For AAV-mediated anterograde labeling, AAV2/1 particles expressing fluorescent protein were injected at P0 into specific cortical sub-regions under guidance by ultrasound backscatter microscopy, as described (Sahni et al., 2021). The central landmarks for the intracranial injections that provide both accuracy and precision are the midline, dorsal and lateral aspects of the lateral ventricle, anterior aspect of the hippocampus, posterolateral aspect of the striatum, corpus callosum, and its genu. All virus work was approved by the Harvard Committee on Microbiological Safety, and conducted according to institutional guidelines.

Klhl14 shRNA constructs

We obtained shRNAmir sequences, which are based on a microRNA-based RNAi (cloned in the pGIPZ lentiviral vector) specifically developed to highly effectively knock down Klhl14 expression (Open Biosystems). In these plasmids, a CMV promoter drives the shRNA and an EGFP reporter. We tested the efficacy of each shRNA in vitro using the psi-CHECK2 luciferase system (Promega, Madison, WI) in 293T HEK cells using the manufacturer’s instructions. The Klhl14 coding sequence was cloned into the psi-CHECK2 vector (Promega, Madison, WI) downstream of the STOP codon in the renilla luciferase gene. Luciferase activity was measured using the Dual-Luciferase Reporter 1000 Assay System (Promega) on a Victor3 1420 plate reader (Perkin Elmer, Waltham, Massachusetts). Efficacy was assessed by comparing luciferase activity of each hairpin with a scrambled shRNA control. The following hairpin sequence gave the strongest knock down (> 85%):

5′-TGCTGTTGACAGTGAGCGACCTGTGTACCCTACAACAAATTAGTGAAGCCA

CAGATGTAATTTGTTGTAGGGTACACAGGCTGCCTACTGCCTCGGA-3′ (mature hairpin sequence: 5′-CTGTGTACCCTACAA-CAAA —3′). This sequence targets the 3′ end of the Klhl14 cDNA just prior to the STOP codon. We subcloned this hairpin into a separate plasmid, where it was placed 3′ to an EGFP coding sequence, driven by a strong chicken beta actin (CBA) promoter (since the CMV promoter is known to be silenced in neural progenitors), and re-confirmed the efficacy in this new backbone. This CBA-driven plasmid was used in the in vivo experiments.

In utero electroporation

Surgeries were performed as previously described (Greig et al., 2016; Molyneaux et al., 2005). For Klhl14 shRNA experiments, the following plasmids were used: pCBA-EGFP-Klhl14 shRNA; pCBA-EGFP-scrambled shRNA (schematized in Figure 5A). For Crim1 overexpression, Crim1 cDNA was placed 3′ to the EGFP coding sequence, driven by the CBA promoter, with the two ORFs separated by a t2A linker sequence. In the control plasmid, Crim1 cDNA was replaced with a STOP codon 3′ to the t2A linker sequence. 2 μg of either plasmid was diluted in PBS for electroporation.

Generation of AAV particles

Adeno associated viral (AAV) 2/1 particles were generated as described (Sahni et al., 2021). Viral particles were generated at the Massachusetts General Hospital Virus Core using established protocols (Maguire et al., 2013). Enhanced green fluorescent protein (EGFP) (for AAV-EGFP), and Crim1 (for AAV-Crim1) coding sequences were cloned into a shuttle plasmid (obtained from the core) that contains the following elements flanked by AAV2 ITRs: a CMV/β-actin promoter to drive the expression of the gene of interest, followed by the woodchuck hepatitis virus post-transcriptional regulatory element (WPRE), a bovine GH pA signal, and an SV40 pA signal. For Crim1 overexpression, AAV-Crim1 particles were co-injected with AAV-EGFP particles (independent experiments with AAV-EGFP and AAV-tdTomato particles confirmed > 95% co-infection efficiency upon co-injection (data not shown)).

For Klhl14 overexpression, Klhl14 cDNA was placed 3′ to the EGFP coding sequence; the two ORFs were separated by a t2A linker sequence. In this plasmid, WPRE was removed in order to include the entire construct under 4.1 kb (between the 2 ITRs). For both Crim1 and Klhl14 overexpression constructs, plasmids were transfected into HEK293T cells, and overexpression was confirmed in vitro using western blot analyses.

AAV8 hsyn-GFP-Cre was obtained from the vector core at the University of North Carolina at Chapel Hill (UNC Vector Core). AAV2/1 CAG-FLEX-tdTomato (originally generated by the Allen Institute) was obtained from the vector core at the University of Pennsylvania. Both viral particles were co-injected into lateral cortex as schematized in Figure 6A.

Immunocytochemistry and in situ hybridization

Brains were fixed and stained using standard methods (Galazo et al., 2016; Greig et al., 2016; Woodworth et al., 2016). Primary antibodies and dilutions used: rat anti-CTIP2, 1:500 (Abcam); rabbit anti-FOG2, 1:250 (Santa Cruz); chicken anti-GFP, 1:500 (Invitrogen); rabbit anti-GFP, 1:500 (Invitrogen); 1:500 rabbit anti-RFP (Rockland Immunochemicals), mouse anti-SATB2, 1:500 (Abcam). In situ hybridization was performed as previously described(Galazo et al., 2016; Greig et al., 2016; Woodworth et al., 2016). cDNA clones for riboprobes are listed in Table S1.

Imaging and quantification

Analysis of retrogradely labeled CSN comparing Crim1 WT and null mice was performed as described in the accompanying study (Sahni et al., 2021). For analyzing CSN retrogradely labeled from cervical C1-C2, 50 μm coronal brain sections of retrogradely labeled brains were imaged on an Axioscan Z1. For cell counts of retrogradely labeled CSN (from cervical cord) in Crim1 null mice, we matched coronal sections of labeled cortices at 4 a priori designated levels (rostral to caudal as in Sahni et al. [2021]; images in Figure 1) in WT, het, and homozygous null mice. Cortices were binned into 5 medio-lateral bins spanning the width of each cortical hemisphere, and medial versus lateral distinction was made by combining the 3 medial bins for medial CSN counts and the 2 lateral bins for lateral CSN counts (as in Sahni et al., 2021). For counts of retrogradely labeled CSNL (for retrograde label injection from caudal thoracic cord), we first dissected the spinal cord to confirm the matched spinal level of the retrograde label injection. Following confirmation, we analyzed coronal sections of each labeled cortex at three distinct rostro-caudal levels selected by a priori criteria (images shown in Figure 2). In all mice, regardless of genotype, labeled CSNL were located only in medial cortex. Matched sections were imaged, and every labeled neuron was counted in each section, using the Cell Counter function in ImageJ.

For unbiased tracing of EGFP+ CSNBC-lat axons in the dorsal funiculus (Figures 4N and 4O), we used the thresholding function in NIH ImageJ. Thresholding criteria were established a priori and then applied to the dorsal funiculus in each monochrome image of the horizontal section of the thoracic spinal cord.

For all CST quantification on axial sections, 60X confocal Z stacks of the entire CST in the dorsal funiculus were obtained on either a BioRad Radiance 2000 or Zeiss LSM 780 confocal microscope. Cervical, thoracic, and lumbar cord axial sections were imaged using identical parameters.

For axon extension experiments, the thoracic cord was sectioned either sagittally or horizontally. Every section that contained a labeled axon was imaged using an ANDOR Clara DR328G camera (ANDOR Technology, South Windsor, CT) mounted on a Nikon Eclipse 90i microscope (Nikon Instruments). Each section was imaged in its entirety, from rostral to caudal and in the Z axis. These Z stacks were collapsed using the ‘‘create focused image’’ function on the NIS-Elements acquisition software (Nikon Instruments), and converted to monochrome images.

For axon counts in in Klhl14-shRNA experiments, we analyzed 3 axial sections each at cervical and thoracic levels and averaged the axon counts. For all fluorescence intensity quantification on axial sections we imaged 2 sections at each spinal level per mouse. We selected the dorsal funiculus as the region of interest (ROI) on all images in Adobe Photoshop. Fluorescence intensity of the labeled CST was then measured in each ROI in monochrome images using ImageJ, and the intensity was averaged at each spinal level. CST fluorescence was normalized to intensity measured at cervical C1, and expressed as a percentage at thoracic and lumbar levels.

For axon extension measurements on horizontal spinal cord sections, we selected 610 × 640 pixel sections of each image of each spinal cord section, at five a priori designated intervals (rostral-most, 25%, 50%, 75%, and caudal most). In each image, the CST was then additionally selected as an ROI in Adobe Photoshop. For all intensity measurements, images first had threshold adjustment using the Feature J function on ImageJ, such that intensity on unlabeled parts of the spinal white matter (without a labeled axon) was zero. Intensity of the ROI was then quantified using the Analyze Particles function on ImageJ. Intensity at each rostro-caudal level was integrated across multiple sections, then normalized to the intensity at the rostral-most limit of the thoracic cord.

Quantification of fluorescence in cortex to evaluate either AAV injection or shRNA electroporation areas was performed as previously described (Greig et al., 2016). Briefly, matched coronal 50 μm thick sections from Klhl14WT and Klhl14flox/flox brains (n = 4 mice for each genotype), all with comparable AAV injections, were imaged for tdTomato on the injected side. Similarly, matched coronal sections from scrambled and Klhl14 shRNA electroporated brains (n = 4 mice for each group), all with comparable electroporations, were imaged for EGFP. In order to quantify fluorescence across the entire mediolateral extent of labeled cortical hemisphere, the straighten plug-in for ImageJ (Kocsis et al., 1991) was first used to straighten the cortical hemisphere. The resulting rectangular image was binned into 100 segments, using the divide slice function in Adobe Photoshop, and fluorescence intensity in each bin was quantified using the measure function in ImageJ. A plot of normalized average fluorescence intensity (and standard error of the mean) against mediolateral position across all brains was generated, and the difference between means for each group was analyzed using an unpaired Student’s t test.

For all experiments, the person analyzing images was blinded to the experimental conditions.

QUANTIFICATION AND STATISTICAL ANALYSIS

Details of imaging quantification methodologies are described in Method Details. All n values and p values are also listed in figure legends. GraphPad Prism version 8 was used to perform statistical tests.

Data distributions were assumed to be normal, but this was not formally tested. For axon counts in Klhl14 shRNA experiments, as well for analysis of normalized CST fluorescence intensity quantification in Klhl14 conditional deletion experiments we used a two-sided t test. We used a two-way ANOVA with repeated-measures followed by Fisher’s least significance difference posthoc test for the axon extension analyses in Crim1GCE mice. For the comparison of three groups in Crim1GCE mice, we used a one-way ANOVA followed by Tukey’s honest significant difference (HSD) posthoc test. No statistical methods were used to pre-determine sample sizes. Variance between groups was analyzed using the f-test procedure.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit anti-CTIP2	Abcam	Cat# ab28448; RRID:AB_1140055
Rabbit anti-Fog2	Santa Cruz Biotechnology	Cat# sc-10755; RRID:AB_2218978
Mouse anti-SATB2	Abcam	Cat# ab51502; RRID:AB_882455
Rabbit anti-GFP	Molecular Probes	Cat# A-1112; RRID:AB_221569
Chicken anti-GFP	Thermo Fisher Scientific	Cat# A10262; RRID:AB_2534023
Rabbit anti-RFP	Rockland	Cat# 600-401-379; RRID:AB_2209751
Bacterial and virus strains
AAV-8 hsyn-GFP-Cre	UNC vector core	N/A
AAV-2/1 FleX-TdTomato	University of Pennsylvania (now at Addgene)	N/A
AAV2/1 CAG-Crim1	This paper	N/A
AAV2/1 CAG-Klhl14	This paper	N/A
AAV2/1 CAG-EGFP	vector core at Massachusetts General Hospital, Boston, MA	Maguire et al., 2013
Chemicals, peptides, and recombinant proteins
Cholera Toxin B subunit, Alexa 555 conjugate	ThermoFisher	Cat # C34776
Critical commercial assays
Dual-Luciferase Reporter 1000 Assay system	Promega	Cat # E1980
Experimental models: Organisms/strains
Crim1GCE mice	(mice obtained from Jackson laboratories) Stock number: 017495	GUDMAP database (Harding et al.; 2011)
Klhl14 flox mice	Embryonic stem cell clones obtained from KOMP database; mice generated in this paper	Skarnes et al., 2011
Oligonucleotides
See Table S1 for oligonucleotide information		N/A
Recombinant DNA
pCAG-GFP-T2A-Klhl14	This paper	N/A
pCAG-Crim1	This paper	N/A
pGIPZ-Klhl14 -shRNAmir clones	Open Biosystems	N/A
Software and algorithms
ImageJ	NIH	https://imagej.nih.gov/ij/
GraphPad Prism 8.0	GraphPad	https://www.graphpad.com/scientific-software/prism/