Transcriptional control of motor pool formation and motor circuit connectivity by the LIM-HD protein Isl2

  1. Yunjeong Lee
  2. In Seo Yeo
  3. Namhee Kim
  4. Dong-Keun Lee
  5. Kyung-Tai Kim
  6. Jiyoung Yoon
  7. Jawoon Yi
  8. Young Bin Hong
  9. Byung-Ok Choi
  10. Yoichi Kosodo
  11. Daesoo Kim
  12. Jihwan Park
  13. Mi-Ryoung Song  Is a corresponding author
  1. School of Life Sciences, Gwangju Institute of Science and Technology, Oryong-dong, Buk-gu, Republic of Korea
  2. Fermentation Regulation Technology Research Group, World Institute of Kimchi, Republic of Korea
  3. Jeonbuk Department of Inhalation Research, Korea Institute of Toxicology, Republic of Korea
  4. Department of Neurology, Samsung Medical Center, Sungkyunkwan University School of Medicine, Republic of Korea
  5. Department of Biochemistry, College of Medicine, Dong-A University, Republic of Korea
  6. Korea Brain Research Institute, Republic of Korea
  7. Department of Brain and Cognitive Sciences, Korea Advanced Institute of Science and Technology (KAIST), Republic of Korea

Abstract

The fidelity of motor control requires the precise positional arrangement of motor pools and the establishment of synaptic connections between them. During neural development in the spinal cord, motor nerves project to specific target muscles and receive proprioceptive input from these muscles via the sensorimotor circuit. LIM-homeodomain transcription factors are known to play a crucial role in successively restricting specific motor neuronal fates. However, their exact contribution to limb-based motor pools and locomotor circuits has not been fully understood. To address this, we conducted an investigation into the role of Isl2, a LIM-homeodomain transcription factor, in motor pool organization. We found that deletion of Isl2 led to the dispersion of motor pools, primarily affecting the median motor column (MMC) and lateral motor column (LMC) populations. Additionally, hindlimb motor pools lacked Etv4 expression, and we observed reduced terminal axon branching and disorganized neuromuscular junctions in Isl2-deficient mice. Furthermore, we performed transcriptomic analysis on the spinal cords of Isl2-deficient mice and identified a variety of downregulated genes associated with motor neuron (MN) differentiation, axon development, and synapse organization in hindlimb motor pools. As a consequence of these disruptions, sensorimotor connectivity and hindlimb locomotion were impaired in Isl2-deficient mice. Taken together, our findings highlight the critical role of Isl2 in organizing motor pool position and sensorimotor circuits in hindlimb motor pools. This research provides valuable insights into the molecular mechanisms governing motor control and its potential implications for understanding motor-related disorders in humans.

Editor's evaluation

This paper will be of interest to developmental biologists who study the gene regulatory mechanisms necessary for neuronal identity and circuit assembly. The study presents important findings regarding the role of the LIM homeodomain transcription factor Isl2 in the development of spinal motor neurons. While the importance of Isl2 for the acquisition of axial and visceral motor neuron development was already described in the literature, the data convincingly describe an additional role in the differentiation of a subset of limb-innervating motor neurons.

https://doi.org/10.7554/eLife.84596.sa0

Introduction

The establishment of motor circuits in the spinal cord involves complex processes including the organization of motor neurons (MNs) into specific groups known as motor columns and motor pools along the rostrocaudal axis (Romanes, 1964; Vanderhorst and Holstege, 1997). In each columnar group, MNs are further segregated into distinct motor pools. The development of this diverse repertoire of motor pools involves the sequentially activated various genetic programs during spinal cord development. Combinatorial gene expression of LIM-homeodomain (HD) transcription factors determines MN identity and positional information across multiple segments, including the median motor column (MMC), the hypaxial motor column (HMC), the lateral motor column (LMC), and the preganglionic column (PGC) (Pfaff et al., 1996; Tsuchida et al., 1994). Hox genes, crucial for body patterning, also contribute to segmental identity by selectively expressing in postmitotic MNs. The absence or misexpression of Hox genes such as Hox6 and Hox10 can lead to disorganization and misdirected targeting of LMC MNs toward specific limb muscles (de la Cruz et al., 1999; Wu et al., 2008). Foxp1, which interacts with Hox genes, is also essential for determining MN fates in the LMC, whose absence leads to scattered and misspecified LMC neurons as demonstrated in Foxp1 mutant mice (Dasen et al., 2008). Nevertheless, despite the involvement of Hox networks, the intricate three-dimensional organization of individual motor pools may require additional regulatory mechanisms, as elongated motor pools can span several spinal cord segments.

More than 50 limb muscles in tetrapods are innervated by distinct limb motor pools, suggesting the involvement of diverse molecular mechanisms in shaping individual limb motor pools during spinal cord development (Sullivan, 1962). Specific LMC motor pools express ETS factors Etv4 and Etv1, playing a role in acquiring motor pool identities. In Etv4 mutant mice, the position and terminal arborization of specific motor pools were perturbed, resulting in disorganized motor pools and impaired motor control (Livet et al., 2002; Vrieseling and Arber, 2006). Gdnf is suggested to act as a peripheral signal derived from limb tissues, guiding major axon bundles toward the hindlimb (Gould et al., 2008; Kramer et al., 2006). Furthermore, Gdnf signaling appears to induce Etv4 expression in certain motor pools, as deletion of Gdnf, its receptor Ret, or Gfra1 results in extinguished Etv4 expression and affected terminal arborization of motor axons in target muscles (Haase et al., 2002; Livet et al., 2002). However, it remains unclear whether the development of Etv4-expressing motor pools solely depends on Gdnf signaling.

Insulin-related, LIM-homeodomain protein 1 (Isl1) and insulin-related, LIM-homeodomain protein 2 (Isl2) are two closely related LIM-HD transcription factors, sharing approximately 75% of their identity, and both are expressed in postmitotic MNs with overlapping patterns. Isl1 is first detected in all newborn MNs shortly after these neurons exit the mitotic cycle at E9.5 (Pfaff et al., 1996; Thaler et al., 2004), while Isl2 begins its expression slightly later, at E10.5 (Thaler et al., 2004). Genetic studies in mice and zebrafish reveal that Isl1 plays a significant role, as its removal induces defects in motor neuronal fates, axonal navigation, neurotransmitter identity, and electrical excitability (Liang et al., 2011; Moreno and Ribera, 2014; Pfaff et al., 1996; Wolfram et al., 2012). Although Isl1 has been considered more crucial than Isl2, evidence suggests that Isl2 is also essential in MN development. In Isl2-null mice, thoracic MNs are affected, leading to scattered or misspecified MN subsets (Thaler et al., 2004). Similarly, misexpression of the dominant negative form of Isl2 or knockdown (KD) of Isl2 in zebrafish results in mispositioned MN cell bodies and defective axon growth (Hutchinson and Eisen, 2006; Segawa et al., 2001). Moreover, successful rescue experiments using isl2 mRNA in isl1 morphants demonstrate that both Isl1 and Isl2 are equally potent in neurons (Hutchinson and Eisen, 2006). While Isl2 initially coexists with Isl1 in pan-MNs, it later becomes enriched in lateral LMC (LMCl), which lacks Isl1 expression (Thaler et al., 2004). This suggests a possible independent role for Isl2 in LMCl beyond its reported function in thoracic MN subsets. Recent advances in transcriptomic analysis reveal detailed spatiotemporal changes in the gene expression of Islet transcription factor during MN development, indicating persistent expression of Isl2 in postnatal MNs (Amin et al., 2021; Catela et al., 2022; Delile et al., 2019; Rayon et al., 2021). Overall, these findings suggest that the role of Isl2 in motor neuronal development may have been underestimated and warrants further investigation, despite the considerable similarities or redundancies in Islet transcription factor.

In this study, we have demonstrated the crucial role of Isl2 in the development and differentiation of hindlimb motor pools. Isl2 regulates the position of motor neuronal cell bodies and directs the assembly of the motor neuronal circuit. Our findings indicate that Isl2 is enriched in proximal motor pools expressing Etv4, and it is involved in neuromuscular junction (NMJ) formation, axonal and dendritic arborization, and hindlimb movement. Through transcriptomic analysis, we have identified Isl2 as a master regulator of various genes involved in the development and differentiation of hindlimb motor pools. Our results indicate that LIM-HD transcription factors, together with Hox and ETS transcription factors, fine-tune the organization of hindlimb motor pools.

Results

Spatiotemporal expression of Islet transcription factors during MN development

Islet transcription factors Isl1 and Isl2 are expressed immediately after MN generation. However, the detailed expression of these factors during MN development at the cellular level remains uncertain. To address this, we re-analyzed published single-cell RNA-sequencing data obtained from E12 MNs, sorted from the cervical to lumbar regions of the spinal cord, focusing on changes in their expression levels (Amin et al., 2021). We categorized approximately 3483 developing MNs into eight subgroups representing major motor columns, such as pMN, immature, MMC, phrenic/hypaxial motor column (P/HMC), LMCm, LMCl, PGCa, and PGCb (Figure 1a and c). Uniform Manifold Approximation and Projection (UMAP) analysis revealed the progression of pMN progenitors into immature MNs, MMC, P/HMC, LMCm, LMCl, PGCa, and PGCb MNs. In earlier immature MNs, we observed high Isl1 expression, which slightly decreased in most differentiated MNs, such as MMC, P/HMC, LMCm, and PGCa MNs, but not LMCl MNs, which did not express Isl1. Isl2 expression emerged later than Isl1 in immature MNs, persisting in MMC and P/HMC neurons. Most LMC neurons showed low but even expression of Isl2, while a few PGCa and b cells expressed Isl2 (Figure 1b). At E10.5, during MN generation, Olig2+ pMNs were initially found in the middle part of the ventral spinal cord, co-expressing Lhx3, a marker for pMNs and MMCs (Lee et al., 2005; Thaler et al., 2002; Figure 1d). As these cells migrated, they transiently expressed markers for postmitotic MNs Isl1 and Hb9. Isl2 expression emerged slightly later, found in more lateral regions across all levels analyzed. To analyze the distribution of Isl2 during the MN maturation, we compared the relative position of three populations: Olig2+ pMNs, Olig2+Isl1+ immature MNs, and Isl2-expressing MNs. Contour and density assessments revealed that Isl2 expression began slightly after Isl1 expression in immature MNs and persisted in postmitotic MNs along with Isl1 (Figure 1d). At E12.5, when motor columns became distinct, we identified them based on the expression of motor neuronal markers, including Foxp1 (marker for LMC neurons), Isl1 (marker for LMCm and PGC neurons), Lhx3 (marker for MMC neurons), nNOS (marker for PGC neurons), and Etv4 (marker for some LMC motor pools). Motor pools expressing Etv4 were located in the LMC neurons but not at the thoracic level (Figure 1e). Isl2 exhibited broad expression across all motor columns, unlike other LIM-HD transcription factors that showed specific expression patterns. Overall, while Isl2 is present in differentiated MNs, it is relatively enriched in MMC and LMCl populations, where its expression is more prominent compared to Isl1.

Spatiotemporal dynamics of Isl2 expression and heterogeneity of motor neurons (MNs).

(a) Uniform Manifold Approximation and Projection (UMAP) visualization of scRNA-seq data based on et al.’s study, highlighting distinct MN lineages: pMN progenitors, immature MNs, MMC, P/HMC, LMCm, LMCl, PGCa, and PGCb neurons, each depicted in distinct clusters. (b) UMAP representation showing the differential expression of Isl1 and Isl2, across MN clusters. (c) Heatmap demonstrating the dynamic gene expression profiles of Isl1, Isl2, and other marker genes within MN clusters. (d) Expression of Olig2, Isl1, Isl2, Hb9, and Lhx3 in E10.5 spinal cords. Contour plots and medio-lateral density plots highlighting Olig2+ MN progenitors (green), Olig2+Isl1+ newborn MNs (blue), Isl2+ MNs (pink) across all spinal cord levels. Scale bars, 50 μm. (e) Expression of Isl1, Isl2, Lhx3, Foxp1, nNOS, Etv4, and Hb9::GFP in E12.5 spinal cords. Scale bars, 100 μm.

Settling position of MNs is impaired in the absence of Isl2

Given the broad expression of Isl2 in postmitotic MNs, we examined the organization of motor columns in Isl2-null mice (Thaler et al., 2004). The pan-motor neuronal expression of Isl2 across the multiple segments of the spinal cord suggests that additional motor neuronal populations could also be affected by the absence of Isl2. To explore unknown phenotypes, we conducted further investigations using the mutant mouse line previously generated and analyzed by Thaler et al., 2004. At E13.5, the number of Hb9+Lhx3+ MMC and Foxp1+ LMC neurons remained unchanged at the brachial and lumbar levels (Figure 2a–c). At the thoracic level, the number of MMC and HMC neurons remained unaffected, but we observed a decrease in the number of Isl1+nNOS+ PGC neurons (Figure 2b and c and Supplementary file 2). Isl2-null mice also showed significantly reduced expression of nNOS and pSMAD in PGC neurons, consistent with previous studies (Dasen et al., 2008; Harrison et al., 1999; Figure 2b). Additionally, the overall distribution of MNs was altered in Isl2-null mice (Figure 2a and b). At the limb level, MN cell bodies were slightly scattered, located between the MMC and LMC neurons. At the thoracic level, ectopic cells were dispersed dorsally to the MMC neurons. Ectopic cells at the lumbar level mainly belonged to LMCl MNs, expressing Lhx1+ (58%), Foxp1+ (53%), and Lhx3+ MMC MNs (32%) (Figure 2—figure supplement 1). Our findings align with previous results, including a reduced number of visceral MNs, reduced nNOS expression, and abnormal dorsomedial migration of MNs, reported by Thaler et al., 2004 Our study additionally reveals the appearance of ectopic MNs at the limb levels, primarily consisting of MMC and LMCl cells.

Figure 2 with 1 supplement see all
Disrupted motor column organization in Isl2-null mice.

(a) Spatial representation of motor neuron (MN) subtypes: median motor column (MMC) (gray, Lhx3+Hb9+), lateral motor column (LMC) (red, Foxp1+), and preganglionic column (PGC) (blue, Isl1+ nNOS+) neurons from the images in b. X and y coordinates denote the medial-to-lateral and ventral-to-dorsal axes, respectively, in μm. (b) Expression of Hb9, Lhx3, Foxp1, Isl1, nNOS, and pSMAD in E13.5 mouse embryonic spinal cords. Brackets mark ectopic MNs observed in Isl2-null mice. (c) Quantification of MMC (gray), LMC (red), hypaxial motor column (HMC) (blue), and PGC (yellow) neurons across brachial (C6–C8), lumbar (L2–L4), and thoracic spinal segments. Average MN number per mouse. Box plots illustrate distribution with median (center line), first and third quartiles (box boundaries), and 10th-90th percentiles (whiskers). Data from three mice per group, analyzed using unpaired Student’s t-test; **p<0.01; n.s. indicates not significant. Refer to Supplementary file 2 and source data for detailed statistics. Scale bars, 100 μm.

Figure 2—source data 1

Quantification of motor neuron (MN) subtypes at E13.5 in Isl2 +/- and Isl2-null mice, categorized by brachial, lumbar, and thoracic levels.

https://cdn.elifesciences.org/articles/84596/elife-84596-fig2-data1-v2.xlsx

To further characterize the scattered MNs at the limb level, we examined the organization of motor pools in Isl2null (knockout [KO]) and Isl2 conditional KO (cKO) mice at the brachial level. Isl2 cKO mice were generated by crossing Isl2F/F mice with Isl2 +/-; Olig2Cre mice (Figure 3—figure supplement 1a and b; Dessaud et al., 2007; Kong et al., 2015). In situ hybridization analysis showed specific downregulation of Isl2 transcript levels among MNs (Figure 3—figure supplement 1c). We assessed MNs based on the expression of marker genes for motor columns and pools, such as Hb9, Lhx3, Foxp1, Isl1, Lhx1 (marker for LMCl neurons), Etv4, and Scip (marker for MMC/HMC MNs and some motor pools), in consecutive transverse spinal cord sections (Figure 3a). At the thoracic level, the previous study reported that the organization of the MMC is disrupted in the absence of Isl2 (Thaler et al., 2004). Similarly, we found that MMC neurons were scattered dorsolaterally and located ectopically within LMC MNs or immediately dorsal to motor columns in Isl2 KO and cKO brachial spinal cords (see Figure 2, Figure 3a and b, Figure 3—figure supplements 2 and 3). The distribution of LMCm was relatively spared, whereas LMCl neurons spread more medially when determined by Foxp1, Isl1, and Lhx1 immunoreactivity. The distribution of major brachial motor pools, defined by Etv4 and Scip expression, remained unchanged in Isl2 mutant mice. At the C6 and C8 levels, the position of major motor pools such as cutaneous maximus (CM), latissimus dorsi (LD), and flexor carpi ulnaris (FCU) motor pools was unaffected in Isl2 KO and cKO mice (Figure 3a, Figure 3—figure supplements 2 and 3; Dasen et al., 2005; Livet et al., 2002). Consequently, the number of individual motor pools also remained unchanged in Isl2 mutant mice (Figure 3c). Thus, at the brachial level, MMC and LMCl neurons were mostly scattered, while the position of other populations remained relatively normal.

Figure 3 with 3 supplements see all
Distribution of lateral motor column (LMC) motor pools in brachial spinal cords of control and Isl2 mutant mice.

(a) Immunofluorescence images of motor neurons (MNs) labeled with Hb9, Lhx3, Foxp1, Scip, Isl1, Lhx1, and Etv4 in adjacent sections of C6 and C8 spinal cords from Het (+/-), Isl2 knockout (KO) (-/-), and Isl2 conditional knockout (cKO) (Isl2 F/-; Olig2Cre) mice. (b) Average ventro-dorsal distribution of MMC (Hb9+Lhx3+) and MMC/HMC (Hb9+Foxp1-Scip+) neurons, and medio-lateral distribution of LMCm (Hb9lowIsl1+) and LMCl (Hb9highLhx1+) neurons in embryos with each genotype. n=3 mice for each genotype; standard deviation (SD) is shown; one-way ANOVA with Bonferroni’s post hoc test; Het vs. KO or cKO, **p<0.01, *p<0.05. (c) Average cell count per embryo in flexor carpi ulnaris (FCU) (Foxp1+Scip+), cutaneous maximus (CM) (Isl1+Etv4+), and latissimus dorsi (LD) (Isl1Etv4+) motor pools. Box plots illustrate distribution with median (center line), first and third quartiles (box boundaries), and 10th-90th percentiles (whiskers). One-way ANOVA with Bonferroni’s post hoc test; n.s. indicates not significant. See Supplementary file 2 and source data for detailed n and statistics. Scale bar, 100 μm.

Figure 3—source data 1

Average ventro-dorsal distribution of MMC (Hb9+Lhx3+) and MMC/HMC (Hb9+Foxp1-Scip+) neurons as well as medio-lateral distribution of LMCm (Hb9lowIsl1+) and LMCl (Hb9highLhx1+) neurons (Figure 3b).

It also includes the average cell count per embryo in flexor carpi ulnaris (FCU) (Foxp1+Scip+), cutaneous maximus (CM) (Isl1+Etv4+), and latissimus dorsi (LD) (Isl1-Etv4+) motor pools (Figure 3c).

https://cdn.elifesciences.org/articles/84596/elife-84596-fig3-data1-v2.xlsx

Next, we reconstructed the detailed motor pool distribution at L2 and L4 levels by analyzing the expression of motor neuronal markers, Hb9, Lhx3, Foxp1, Scip (marker for MMC/HMC neurons), Nkx6.1 (marker for MMC and LMCm neurons), Lhx1, Isl1, and Etv4 (Figure 4). Scip+ MMC/HMC neurons were scattered dorsally, and Nkx6.1+ LMCm and Lhx1+ LMCl neurons were scattered medially in both Isl2 KO and cKO mice, as shown in contour, density, or spatial plots (Figure 4a and b, Figure 4—figure supplements 1 and 2). Lhx1+ LMCl neurons were scattered medially, while Nkx6.1+ LMCm neurons were less distinct in Isl2 mutant mice (Figure 4a, b, and c and Figure 4—figure supplements 1 and 2). The clustering of tibialis anterior (Ta) motor pools was relatively normal in Isl2 mutant mice (Figure 4c). Remarkably, the number of Etv4-expressing cells, including gluteus (Gl), rectus femoris (Rf), and tensor fasciae latae (Tfl) motor pools in the L2 and L4 segments, was significantly downregulated, as shown in contour and spatial plots (Figure 4a, c, and d and Figure 4—figure supplements 1 and 2). Taken together, MMC neurons at all levels and some LMC motor pools at the lumbar level were dispersed in the absence of Isl2.

Figure 4 with 2 supplements see all
Altered lumbar motor column position in Isl2 mutant mice.

(a) Immunofluorescence images of motor neurons (MNs) labeled with Hb9, Lhx3, Foxp1, Scip, Nkx6.1, Lhx1, Isl1, and Etv4 in adjacent sections of L2 and L4 spinal cords from Het (+/-), Isl2 knockout (KO) (-/-), and Isl2 conditional KO (cKO) (Isl2 F/-; Olig2Cre) mice. (b) Average ventro-dorsal distribution of MMC (Hb9+Lhx3+) and MMC/HMC (Hb9+Foxp1-Scip+) neurons and medio-lateral distribution of LMCm (Hb9lowNkx6.1+) and LMCl (Hb9highLhx1+) neurons in heterozygote, Isl2 KO, and Isl2 cKO embryos. n=3 mice for each genotype; SD is shown; one-way ANOVA with Bonferroni’s post hoc test; **p<0.01, *p<0.05. See Supplementary file 2 and source data for detailed n and statistics. (c) The average cell count per embryo in Etv4+ rectus femoris (Rf)/tensor fasciae latae (Tfl)/gluteus (Gl) and Nkx6.1+Lhx1+ Ta motor pools. n=3 mice for each genotype; SEM is shown; one-way ANOVA with Bonferroni’s post hoc test, ***p<0.001, n.s.=not significant. Box plots illustrate data distribution with median (center line), first and third quartiles (box boundaries), and 10th-90th percentiles (whiskers). See Supplementary file 2 and source data for detailed n and statistics. (d) Summary diagram depicting the position of major lumbar motor pools: rectus femoris (R), tensor fascia latae (T), tibialis anterior (Ta), and Gl. Misspecified motor pools are marked with asterisks. Scale bar, 100 μm.

Figure 4—source data 1

Average ventro-dorsal distribution of MMC (Hb9+Lhx3+) and MMC/HMC (Hb9+Foxp1-Scip+) neurons, as well as the medio-lateral distribution of LMCm (Hb9lowNkx6.1+) and LMCl (Hb9highLhx1+) neurons (Figure 4b).

It also includes the average cell count per embryo in Etv4+ rectus femoris (Rf)/tensor fasciae latae (Tfl)/gluteus (Gl) and Nkx6.1+ Lhx1+ tibialis anterior (Ta) motor pools (Figure 4c).

https://cdn.elifesciences.org/articles/84596/elife-84596-fig4-data1-v2.xlsx

Isl2 induces Etv4 transcripts in lumbar motor pools

The presence of scattered MNs and the absence of Etv4 expression in Isl2 mutant lumbar motor pools led us to hypothesize that Isl2 activates Etv4 expression, as its absence also resulted in positional defects in MNs (Livet et al., 2002). To investigate this, we examined Etv4 expression in Isl2 mutant mice during the segregation of limb motor pools from E11.5 to E13.5. Immunohistochemistry and RT-PCR analysis using spinal cord tissues revealed that both the proteins and transcripts of Etv4 were selectively downregulated in lumbar motor pools, but not in brachial motor pools, of Isl2 KO and Isl2 cKO spinal cords (Figure 5a and b and Figure 5—figure supplement 1).

Figure 5 with 2 supplements see all
Loss of Etv4 expression in Isl2-null motor neurons (MNs).

(a) Expression of Isl1, Scip, and Etv4 in brachial and lumbar spinal cords of Isl2 +/-, Isl2 knockout (KO), Isl2 F/−, and Isl2 F/−; Olig2Cre animals at E11.5 and E13.5. Note that Etv4 expression vanishes in lumbar, but not in brachial MNs (arrowheads). Summary diagram depicts the position of lateral motor column (LMC) motor pools innervating the cutaneous maximus (CM), latissimus dorsi (LD), and gluteus (Gl) muscles for each genotype. Scale bars, 100 μm. (b) RT-PCR results and quantification demonstrate reduced Etv4 transcript levels in E12.5 Isl2-null lumbar spinal cords. Relative expression of Etv4 at brachial: 1.00±0.06 for Isl2 +/+, 0.98±0.04 for Isl2 +/-, and 0.85±0.07 for Isl2 KO; p=0.2692 for Isl2 +/+ vs. Isl2 KO, p=0.3736 for Isl2 +/- vs. Isl2 KO; n=3 animals per genotype. Relative expression of Etv4 at lumbar: 1.00±0.08 for Isl2 +/+, 0.96±0.08 for Isl2 +/-, and 0.54±0.07 for Isl2 KO; p=0.0020 for Isl2 +/+ vs. Isl2 KO, p=0.0024 for Isl2 +/- vs. Isl2 KO; n=6–8 animals per genotype. Box plots illustrate distribution with median (center line), first and third quartiles (box boundaries), and 10th-90th percentiles (whiskers). One-way ANOVA with Tukey’s test; see Supplementary file 2 and source data for detailed n and statistics. **p<0.01, n.s.=not significant. (c) Expression of GFP, MNR2, ISL2, ALDH1A2, and ETV4 in HH stage 25 chick neural tubes, electroporated with siRNA against ISL2. Quantification analysis includes the average MNR2-expressing MNs per embryo (96±17 MNs for control, vs. 95±19 for siRNA-electroporated chick, p=0.9622), average fluorescent intensity of ISL2 expression per embryo (1.00 of relative expression ratio for control, vs. 0.65±0.06 for siRNA-electroporated side, p=0.0008), and average relative expression of ALDH1A2 per embryo (1.00 for control, vs. 0.85±0.10 for siRNA-electroporated side, p=0.3340), and average relative expression of ETV4 per embryo (1.00 for control, vs. 0.13±0.01 for siRNA-electroporated side, p<0.0001). Note that the ETV4 transcript was downregulated on the electroporated side (right; n=3–4 animals per group; SEM is shown; unpaired Student’s t-test; see Supplementary file 2 and source data for detailed n and statistics). ****p<0.0001, ***p<0.001, n.s.=not significant. Scale bars, 50 μm.

Figure 5—source data 1

Quantification of relative expression of Etv4 in brachial and lumbar spinal cords of wild-type, Isl2 +/- and Isl2-null embryos for Figure 5b.

https://cdn.elifesciences.org/articles/84596/elife-84596-fig5-data1-v2.xlsx
Figure 5—source data 2

Quantification of MNR2+ cells and relative expression of ISL2, ETV4, ALDH1A2 in chick spinal cords, electroporated with siRNA against ISL2 for Figure 5c.

https://cdn.elifesciences.org/articles/84596/elife-84596-fig5-data2-v2.xlsx
Figure 5—source data 3

Unedited raw agarose gel pictures and labeled gel pictures of RT-PCR results for Figure 5b.

https://cdn.elifesciences.org/articles/84596/elife-84596-fig5-data3-v2.zip

To further test whether acute downregulation of Isl2 is sufficient to abolish Etv4 transcription, we knocked down ISL2 expression in the chick neural tube using siRNA. Efficient KD of ISL2 was confirmed in a cell line misexpressing HA-tagged ckISL2 (Figure 5—figure supplement 2). When ISL2 siRNA was introduced into the chick neural tube by in ovo electroporation, the number of ISL2-expressing cells and the intensity of ISL2 in these cells were reduced, while the total number of MNs labeled with MNR2, a marker expressed among MNs and progenitors in chicken embryos, remained unchanged (Tanabe et al., 1998; Figure 5c). Consistent with the results in mice, the ETV4 transcript level was significantly downregulated in chick ISL2-KD MNs (Figure 5c). Collectively, these results suggest that Isl2 is necessary for gene transcription of Etv4 in MNs in a cell-autonomous manner, which may correlate with the correct positioning of motor pools.

Transcriptomic analysis to define genes is associated with the development of lumbar motor pools

Due to the more severe effect observed in the absence of Isl2 on lumbar motor pools at the lumbar level compared to other axial levels, we sought to identify sets of genes under the control of Isl2, particularly in the lumbar segment. We applied two search criteria. First, we focused on downregulated genes, as Isl2 is expected to act as an activator similar to Isl1 (Lee et al., 2008; Lee and Pfaff, 2003). Second, we compared differentially expressed genes (DEGs) at the brachial and lumbar levels to select genes specifically downregulated at the lumbar level. To achieve this, we conducted bulk RNA-seq analysis using E12.5 brachial and lumbar ventral spinal cords. To obtain MN-enriched tissues, we dissected the ventral spinal cord from an open-book preparation. Segmental identity of the spinal cord was verified according to the expression of the correct Hox code at each axial level (Hayashi et al., 2018). The DEG analysis between Isl2 KO and Isl2 heterozygote mice revealed 140 genes downregulated at the brachial levels and 159 genes downregulated at the lumbar levels (<−0.5 log2-FC, raw p<0.05). We then identified DEGs that overlapped with the list of genes from the scRNA-seq dataset obtained from embryonic MNs to further identify MN-specific genes (Amin et al., 2021). As a result, 44 and 87 DEGs were specifically downregulated at the brachial and lumbar levels of Isl2 KO spinal cords, respectively, whereas nine DEGs were commonly downregulated in both segments (Figure 6a–e and Supplementary file 3a-b). Notably, gene ontology (GO) analysis revealed that the downregulated genes at the lumbar level in Isl2-null embryos were predominantly associated with the neuropeptide signaling pathway, synaptic signaling, locomotion, neuron differentiation, and axon development (Figure 6g). In contrast, at the brachial level, no categories involving synapse organization, axon development, and neuron differentiation were detected (Figure 6f). To validate some of these findings, we investigated the top 20 DEGs (p-value <0.001) and confirmed their expression in Isl2 mutant spinal cords (Figure 6h). For instance, A730046J19Rik transcript was detected in Rf motor pools at the L2 level of E12.5 spinal cords and was downregulated in Isl2 KO. Expression of Anxa2, Kcnab1, Etv4, Prph, and C1ql3 was also found in Gl motor pools, the most ventrolaterally located motor pool subsets, at the L4 level and was markedly downregulated in the Isl2 KO spinal cords, consistent with normalized read count values.

Figure 6 with 1 supplement see all
Transcriptome analysis of Isl2-deficient brachial and lumbar motor neurons (MNs) at E12.5.

(a, b) MA plots highlighting top differentially expressed genes (DEGs) in brachial and lumbar MNs of E12.5 Isl2-null and Isl2 +/- embryos. (c) Venn diagram depicting the overlap of downregulated genes in brachial and lumbar MNs in Isl2-null embryos. (d, e) Heatmaps illustrating selected downregulated genes in the brachial and lumbar spinal cords of Isl2-deleted embryos. (f, g) Gene ontology (GO) analysis of downregulated DEGs at the brachial and lumbar levels in Isl2 knockout (KO) embryos. (h) In situ hybridization and immunohistochemistry of selected downregulated genes in E12.5 Isl2-deficient lumbar spinal cords. Dotted lines indicate the position of LMCm and LMCl. Data points of the graphs indicate normalized read counts from individual animals. SEM is shown. (i) Expression assessment of Epha3, Fgf10, and Sema5a, genes enriched in sb.LMCl.2 subcluster from scRNA-seq analysis in Figure 6—figure supplement 1. Scale bars, 50 μm.

To pinpoint the specific hindlimb motor pools affected by the downregulation of Isl2, we further analyzed the LMC motor pools using a published single-cell RNA-sequencing dataset from E12.5 mouse spinal cords (Amin et al., 2021). We isolated the LMC subclusters in the rostral lumbar spinal cord (L1 to L4) that expressed high levels of Hoxd9, Hoxc10, Hoxd10, Hoxa11, Hoxc11, and Hoxd11 (Figure 6—figure supplement 1). UMAP analysis of 273 cells identified five clades: two LMCl and three LMCm subclusters (Figure 6—figure supplement 1a–h). Subcluster LMCl.1 (sb.LMCl.1) expressed Isl2 and Lhx1 but not Etv4 and was further divided into two subclusters: sb.LMCl.1.v, consisting of vasti (V) motor pools expressing Isl2, Lhx1, and Etv1; and sb.LMCl.1.ta, consisting of Ta motor pools expressing Isl2, Lhx1, and Nkx6-1. sb.LMCl.2 contained Rf, Tfl, and Gl motor pools that expressed Lhx1, Nkx6-2, and Etv4. sb.LMCm.1 was characterized by the expression of Isl2, Isl1, and Etv1, with molecular profiles similar to the adductor/gracilis (A/G) motor pools. Subclusters sb.LMCm.2 and sb.LMCm.3 were both defined by the expression of Isl2, Isl1, and Nkx6-1 and contained hamstring (H) and gastrocnemius (Gs) motor pools, with sb.LMCm.2 (Hoxd9high and Aldh1a2high) occupying a more rostral position than sb.LMCm.3 (Hoxd9low and Aldh1a2low) (Figure 6—figure supplement 1). Thus, we successfully defined five subclusters of LMC motor pools. Among the LMC subclusters, Isl2 expression was higher in sb.LMCl.2 than in other subclusters, suggesting a close genetic relationship between Isl2 and Etv4 (Figure 6—figure supplement 1b-d). Moreover, some genes were relatively enriched in specific motor pools, which is potentially useful for distinguishing individual motor pools. For instance, Epha3, Fgf10, and Sema5a transcripts were enriched in Gl motor pools and their expression was reduced in Isl2 KO, as validated by in situ hybridization analysis (Figure 6i, Supplementary file 3c). In conclusion, our integrated approach, combining bulk and scRNA-seq analyses datasets, has proven valuable in identifying new potential motor pool markers and uncovering novel transcription target genes.

Defective arborization and sensorimotor connectivity of Isl2-deficient hindlimb motor pools

The downregulation of Etv4 and the transcriptomic analysis of Isl2-deficient mice suggested that these mice have similar phenotypes to those found in Etv4 mutant mice, such as scattered cell bodies, reduced dendrite patterning, and sensorimotor connectivity (Vrieseling and Arber, 2006). At the lumbar level of the spinal cord, there were three motor pools that expressed Etv4: Gl, Rf, and Tfl motor pools (Arber et al., 2000; De Marco Garcia and Jessell, 2008). Thus, we sought to investigate whether Isl2 plays a crucial role in dendrite formation of these motor pools. A previous study by Thaler et al., 2004, had established that Isl2 KO mice were unable to survive beyond a day after birth (Thaler et al., 2004). Intriguingly, our study revealed that Isl2 cKO mice also died on the first day after birth, while 11% of Isl2 KO mice managed to survive (Supplementary file 1, Video 1). While the precise underlying cause for this disparity in survival rates among Isl2 KO mutant mouse lines requires further investigation, we were able to examine adult MNs in the absence of Isl2. Overall, the total number of ChAT+ MNs remained unchanged, and mislocalized MNs were observed in the P21 Isl2-null lumbar spinal cord (Figure 7a–c). We next examined the dendrite formation of motor pools by injecting rhodamine-dextran (Rh-Dex) retrograde tracer into individual muscles of Isl2 KO postnatal pups at P4. In the wild-type animals, the Gl motor pools exhibited a stereotyped elongated crescent shape of the dextran-labeled dendritic arbor at the ventrolateral position in L3-L4 spinal cords (Figure 7d). However, in Isl2 KO mice, dextran-labeled Gl motor pools split into two groups: 61% of cells were located at the correct ventrolateral position, and 39% were located at the ectopic ventromedial position. Regardless of position, Isl2-deficient Gl motor pools showed shorter and more randomly oriented dendritic arbors (Figure 7d). Similarly, Tfl motor pools, whose axons diverged from one of the major branches of Gl ones, also displayed shorter and lesser dendritic arbors in Isl2 KO mice (Gould et al., 2008). The position and dendritic arborization of Rf motor pools remained unchanged in Isl2 KO mice (Figure 7d). Considering that Etv4 expression has been previously reported to be transient until E13.5 in Rf motor pools, it may not play a significant role in the dendritic arborization of Rf motor pools in adults (De Marco Garcia and Jessell, 2008). To investigate the connectivity of Gl motor pools further, we next analyzed the density of sensory synaptic contacts in control and Isl2 KO mice. Rh-Dex was injected into the Gl muscle at P16, and the number and density of vesicular glutamate transporter 1 (vGluT1), a marker of proprioceptive synapses, were analyzed at P21. We analyzed the synaptic density in two different regions: zone 1, containing normal LMC motor columns, and zone 2, where ectopic MNs reside (Figure 7e). In control mice, all Rh-Dex+ Gl MNs were located in zone 1, while in Isl2 KO mice, only about 42% were found in zone 1 (Figure 7e–g).

Impaired proprioceptive sensory nerve connectivity in the gluteus (Gl) motor pools of Isl2 mutant mice.

(a–c) Representative images of adult lumbar spinal cords immunostainined with ChAT antibody and quantification of average number of motor neurons (MNs) (b) and their medio-lateral distribution (c). (d) Representative dendritic arbors of Gl, tensor fasciae latae (Tfl), and rectus femoris (Rf) motor pools retrogradely labeled with rhodamine-dextran (Rh-Dex). Radial (R) plots show average dendritic membrane density per octant (red bars) from motor somata (six to eight adjacent sections per group, three animals per group). (e) Representative images of P21 Isl2 +/- and Isl2-null lumbar spinal cords with designated zone 1 and zone 2 regions. (f) Medio-lateral distribution of Rh-Dex-labeled Gl motor pools in Isl2 +/- and Isl2 KO mice. (g) Average medio-lateral position of Gl motor pools in Isl2 +/- and Isl2 knockout mice with box plot depicting the number of dextran-labeled Gl neurons. (h) Quantification of vGluT1 intensity in zone 1 and zone 2 in Isl2 +/- and Isl2 KO mice. (i) Representative images showing the contact of vGluT1+ Gl sensory boutons with Rh-Dex-labeled Gl MNs (arrowheads). (j) The number of sensory boutons per cell body (12.4±1.2 boutons for Isl2 +/-, vs. 6.4±0.8 boutons for Isl2 KO, p=0.0002). (k) The number of sensory boutons per cell body in zone 1 and zone 2 (12.4±1.2 boutons for zone 1 in Isl2 +/-, 9.5±1.0 boutons for zone 1 in Isl2 KO, and 4.2±0.5 for zone 2 in Isl2 KO). p=0.1488 for zone 1 in Isl2 +/- vs. zone 1 in Isl2 KO; p<0.0001 for zone 1 in Isl2 +/- vs. zone 2 in Isl2 KO; p=0.0062 for zone 1 in Isl2 KO vs. zone 2 in Isl2 KO. Box plots illustrate distribution with median (center line), first and third quartiles (box boundaries), and 10th-90th percentiles (whiskers).; unpaired Student’s t-test for b, g, h, j and one-way ANOVA with Tukey’s test for k; see Supplementary file 2 and source data for detailed n and statistics. ****p<0.0001, ***p<0.001, **p<0.01, n.s.=not significant. Scale bars: (a) 100 μm, (d) 100 μm, (e) 100 μm, (i) 10 μm.

Figure 7—source data 1

Average number of ChAT-expressing motor neurons (MNs) for Figure 7b.

https://cdn.elifesciences.org/articles/84596/elife-84596-fig7-data1-v2.xlsx
Figure 7—source data 2

Quantification of the number of vGluT1+ sensory boutons per cell body and intensity in P21 Isl2 +/- and Isl2 knockout (KO) mice for Figure 7f–h and j–k.

https://cdn.elifesciences.org/articles/84596/elife-84596-fig7-data2-v2.xlsx
Video 1
Example movie of a new born littermate Isl2F/+ and an Isl2 conditional knockout (cKO) mouse.

Next, we analyzed the synaptic density of vGluT1+ boutons on motor neuronal somata. Overall, the density of vGluT1 immunoreactivity in zone 1 was higher than in zone 2, which is expected since more MNs reside in zone 1 where more sensorimotor synapses are present. Compared to the control mice, the density of vGluT1 immunoreactivity in Isl2 KO mice did not change, indicating that overall sensorimotor connection is not affected in them (Figure 7h). However, when we analyzed the number of vGluT1+ boutons in contact with soma of Rh-Dex+ Gl MNs, we observed a great reduction in the number of boutons in Isl2 KO mice (Figure 7i–k). Furthermore, the number of boutons in zone 2, but not in zone 1, was significantly reduced in Isl2 KO mice (Figure 7k). Thus, ectopically located Gl MNs are more likely to receive fewer proprioceptive sensory inputs compared to the MNs in a normal position in the absence of Isl2. Collectively, our findings suggest that Isl2 KO mice displayed reduced and misoriented dendrites of Gl motor pools, leading to defective sensorimotor connections.

Aberrant NMJ formation and terminal axon branching in Isl2 KO mice

To investigate whether the axon projections from mispositioned motor columns successfully reach their target muscles in the limb, we traced the trajectories of LMC axons in Isl2 KO mice that carried the Hb9::GFP. LMC axons reach the base of the limb and diverge to either dorsal or ventral limb trajectories. The major brachial and lumbar plexus appeared mostly normal in Isl2 KO mice (Figure 8—figure supplement 1a–c). These results differ from previous studies involving mice with defects in Gdnf ligands or receptors, where the peroneal nerves were severely affected, resulting in limited or no innervation of the target muscles (Bonanomi et al., 2012; Gould et al., 2008; Kramer et al., 2006). Additionally, in our bulk RNA-seq analysis, we found that the expression levels of Gdnf receptors Ret and Gfra1 was unchanged (–1.13-fold, –1.18-fold each), and their transcripts remained detectable in Isl2-deficient Gl motor pools when assessed by in situ hybridization analysis (Figure 8—figure supplement 1d). Thus, these findings suggest that Gdnf-dependent guidance of LMC projections in the hindlimb may not necessarily depend on Isl2.

We then examined the formation of the NMJ in the hindlimb muscles. Due to the postnatal lethality of Isl2 cKO, we analyzed the muscles of these mice at E18.5, prior to birth, when NMJ formation is in progress (Figure 8—figure supplement 2). GFP-labeled axons arrived in Isl2-deficient Gl muscles; however, their terminal arbors were defasciculated and misrouted. Fluorescence intensity of α-bungarotoxin (BTX) at the NMJs was greatly reduced in Isl2 cKO. Furthermore, AChR clustering appeared to be incomplete, showing slender and elongated NMJs, or some NMJs were abnormally concentrated in proximity to the primary axon shaft in Isl2 cKO (Figure 8—figure supplement 2b).

Because NMJs mature during the first postnatal weeks, we next examined the Gl muscles at birth when NMJ formation is ongoing, and at 2 weeks of age when synaptic elimination is almost complete (Wyatt and Balice-Gordon, 2003). At P0, Hb9::GFP-labeled main motor axon bundles were comparable across Isl2-deficient Gl muscles; however, the branches originating from the main bundles and the terminal arbors were drastically diminished. The number of NMJs in Gl was reduced to 84% in the Isl2 mutant group, and the length and complexity of the secondary branches were significantly reduced in P0 (Figure 8a–c). Higher magnification views revealed that numerous short aberrant sprouts and NMJs had developed near the primary branches in Isl2-deficient muscles, which was not observed in the control group (Figure 8a). The end plate band area occupied by motor axons was reduced to 55%, and the end plate band area occupied by AChR clusters marked by BTX was reduced to 58% in the Gl muscles of the control group (Figure 8a, d, and e). At P14, a drastic loss in AChR clusters and excessive growth of motor axons that bypassed the AChR clusters were observed in Isl2 KO mice (Figure 8a, f, and g). When the few surviving NMJs were visualized during the period of synaptic elimination, a significant portion of nerve terminals were aberrant in Isl2 KO mice, showing faint AChR clusters, polyinnervation, denervation, and swelling of axons. The area of the NMJ defined by the nerve terminal and the area of AChR clusters were also enlarged in the muscles of Isl2 KO mice. Fragmented and less compact AChR clusters were more abundant in Isl2 KO mice than in heterozygote mice (Figure 8k–r). In Tfl muscle, abnormal AChRs were formed along axon branches at P0, and a drastic loss in AChR clusters and polyinnervation were observed at P14 and P28 in Isl2 KO mice (Figure 8h, i, and j). However, Rf muscles, whose motor pools transiently express Etv4, showed normal arrangement of NMJs (Figure 8—figure supplement 3; De Marco Garcia and Jessell, 2008). Other muscles innervated by non-Etv4-expressing motor pools had normal NMJ development (Figure 8—figure supplement 3). These findings collectively suggest that the primary role of Isl2 lies in the axon terminal arborization of α-MNs within Etv4-expressing motor pools.

Figure 8 with 3 supplements see all
Reduced terminal motor axon branching in gluteus (Gl) muscles of Isl2 mutants.

(a) Visualization of motor axons and neuromuscular junctions (NMJs) in P0 and P14 Gl muscles with Hb9::GFP (green) and α-bungarotoxin (BTX, red) immunoreactivity. Higher magnification views are provided for the boxed regions. The NMJ region is delineated by white lines in P0. Arrowheads indicate atrophic NMJs on overshooting axons in P14. (b) Number of NMJs in P0 Gl muscles (919±23 NMJs for control, vs. 772±44 for Isl2 knockout [KO], p=0.0257, n=4 muscles per group, three animals). (c) Secondary branch length (313±17 μm for control, vs. 92±8 μm for Isl2 KO, p<0.0001, n=90–120 axons per group, three animals) in P0 Gl muscles. (d) End plate area covered by motor axons (1.66±0.15 mm2 for control, vs. 0.92±0.06 mm2 for Isl2 KO, p=0.0014, n=4–5 muscles per group, three animals) in P0 Gl muscles. (e) End plate area covered by AChR clusters (1.04±0.05 mm2 for control, vs. 0.61±0.05 mm2 for Isl2 KO, p=0.0005, n=4–5 muscles per group, three animals) in P0 Gl muscles. (f) The number of AChR clusters (828±45 NMJs for Isl2 +/-, vs. 104±18 for Isl2 KO, p<0.0001, n=4 muscles per group, three animals) in P14 Gl muscles. (g) Axons extending beyond AChR clusters at P14 (0.26±0.07 axons for Isl2 +/-, vs. 3.92±0.29 for Isl2 KO, p<0.0001, n=39–46 100 μm intervals of end plate per group, three animals). (h) Visualization of motor axons and NMJs in P0, P14, and P28 tensor fasciae latae (Tfl) muscles. (i) The number of NMJs at P14 Tfl muscles (324±35 NMJs for Isl2 +/- vs. 34±5 for Isl2 KO, p=0.0002, n=4 muscles per group, three animals). (j) The number of axonal inputs in P28 Tfl muscles (1.09±0.09 axons for Isl2 +/-, vs. 4.08±0.40 axons for Isl2 KO, p<0.0001, n=11–13 NMJs per group, three animals). (k) Representative examples of NMJs with morphological defects at higher magnification in Gl muscles at P14. For compactness and fragmentation analysis, P28 NMJs were analyzed. A swelling axon is indicated by an arrowhead. (l–r) Percentage of faint NMJs (1.12 ± 0.48% for Isl2 +/-, vs. 70.47 ± 3.22% for Isl2 KO, p<0.0001, n=16–17 random fields per group), polyinnervated NMJs (0.81 ± 0.26% for Isl2 +/-, vs. 18.49 ± 2.75% for Isl2 KO, p<0.0001, n=8–17 random fields per group), number of axonal inputs (1.05±0.05 axons for Isl2 +/-, vs. 3.91±0.25 axons for Isl2 KO, p<0.0001, n=11–22 NMJs per group), area of AChR clusters (357±11 μm2 for Isl2 +/-, vs. 701±57 μm2 for Isl2 KO, p<0.0001, n=11–22 NMJs per group) and nerve terminal (270±10 μm2 for Isl2 +/-, vs. 696±63 μm2 for Isl2 KO, p<0.0001, n=11–22 NMJs per group), percentage of denervated NMJs (0.08 ± 0.08% for Isl2 +/-, vs. 4.95 ± 1.04% for Isl2 KO, p<0.0001, n=16–17 random fields per group), compactness (81.78 ± 1.07% for Isl2 +/-, vs. 55.52 ± 1.99% for Isl2 KO, p<0.0001, n=15–39 NMJs per group) and fragmentation (0.06±0.03 for Isl2 +/-, vs. 0.78±0.06 for Isl2 KO, p<0.0001, n=15–39 NMJs per group) of AChR clusters were measured. n=3 animals per group; box plots illustrate distribution with median (center line), first and third quartiles (box boundaries), and 10th and 90th percentiles (whisker); unpaired Student’s t-test in b-f, and o; Mann-Whitney test in g, j, l-n and p-r; see Supplementary file 2 and source data for detailed n and statistics. ****p<0.0001, ***p<0.001, **p<0.01, *p<0.05, n.s.=not significant. Scale bars: (a) 200 μm (for low-magnification images) and 50 μm (for high-magnification images), (h) 20 μm in P0, 100 μm in P14, and 10 μm in P28. (k) 10 μm.

Figure 8—source data 1

Quantification of gluteus (Gl) neuromuscular junctions (NMJs) in Isl2 +/- and Isl2 knockout (KO) mice for Figure 8b–g and l–r.

https://cdn.elifesciences.org/articles/84596/elife-84596-fig8-data1-v2.xlsx
Figure 8—source data 2

Quantification of tensor fasciae latae (Tfl) neuromuscular junctions (NMJs) in Isl2 +/- and Isl2 knockout (KO) mice for Figure 8i–j.

https://cdn.elifesciences.org/articles/84596/elife-84596-fig8-data2-v2.xlsx

Impaired hindlimb movement in Isl2-deficient mice

To determine whether defective sensorimotor connectivity ultimately caused deficits in hindlimb movements, we examined the behaviors of adult Isl2 mutant mice. We found that about 10% of Isl2 KO mice survived after birth, although they were smaller with lower body weight (Figure 9e, Supplementary file 1). Unexpectedly, however, Isl2 cKO mice, similar to the Isl2 KO mice as previously reported, failed to survive after birth (Supplementary file 1, Video 1; Thaler et al., 2004). Although the discrepancy for the lethality needs to be further investigated, we decided to monitor the postnatal abnormalities using our Isl2 KO mice. Interestingly, newborn Isl2 KO mice exhibited rigid hindlimbs that were parted widely, unlike their forelimbs, during free walking (Video 2). Adult Isl2 KO animals also maintained an unnatural extended hindlimb posture when suspended by the tail and during walking (Figure 9a and Video 3). Footprint analysis showed that Isl2 KO mice had abnormal gait of the hindlimb and broader width than that of the control group, while the gait of the forelimbs was normal (Figure 9b and c). A rigid and broad posture of the hindlimb may be a sign of proximal muscle weakness. Indeed, the muscle mass of the hip muscles, including the Gl and Rf muscles, was significantly reduced in Isl2 KO animals (Figure 9d and e). X-ray imaging and whole skeleton staining with Alizarin red and Alcian blue showed that the overall bone structure and growth were normal in these animals (Figure 9—figure supplement 1a and b). We then conducted electromyographic (EMG) analysis to measure hindlimb muscle activity in Isl2 KO mice. Recordings from the Gl muscles in free-walking Isl2 KO mice showed lower firing frequency, fewer single-motor-unit potentials, and shorter burst activity duration than the heterozygote animals (Figure 9f–j). Taken together, our results show that Isl2 mutant mice displayed abnormal rigid hindlimb movement, which was accompanied by impaired EMG activity.

Figure 9 with 1 supplement see all
Hindlimb gait impairment in Isl2 mutants.

(a) Tail suspension test images showing rigid wide-open posture of the hindlimbs in Isl2 mutant mice. (b) Footprint patterns of 3-month-old Isl2 +/- and Isl2 knockout (KO) mice. (c) Measurement of the stride width of forelimbs (1.56±0.06 cm for Isl2 +/-, vs. 1.73±0.06 cm for Isl2 KO, p=0.0782) and hindlimbs (2.70±0.05 cm for Isl2 +/-, vs. 3.63±0.09 cm for Isl2 KO, p<0.0001). n=5–8 animals per genotype; SEM is shown; unpaired Student’s t-test; see Supplementary file 2 and source data for detailed n and statistics. ****p<0.0001, n.s.=not significant. (d) Gross appearance of lower limbs and dissected hindlimb samples of adult Isl2 +/- and Isl2 KO mice. (e) Measurement of body (9.9±0.8 g for Isl2 +/-, vs. 6.9±0.8 g for Isl2 KO, p=0.0339) and muscle weight for gluteus (GI) (67.6±12.6 mg for Isl2 +/-, vs. 34.3±2.7 mg for Isl2 KO, p=0.0316, n=5–6 muscles per group) and rectus femoris muscles (51.1±8.2 mg for Isl2 +/-, vs. 27.6±3.8 mg for Isl2 KO, p=0.0384, n=5–6 muscles per group). n=5 animals for heterozygote, n=4 animals for KO; SEM is shown; unpaired Student’s t-test; see Supplementary file 2 and source data for detailed n and statistics. *p<0.05. (f) Representative electromyographic (EMG) recordings of Gl muscles during free-walking. (g) Signature of individual footsteps. (h–j) Quantification of average frequency (861±18 Hz for Isl2 +/-, vs. 619±16 Hz for Isl2 KO, p<0.0001), number of single-motor-unit potentials (80±4 peaks for Isl2 +/-, vs. 42±3 peaks for Isl2 KO, p<0.0001), and burst activity duration (0.096±0.005 s for Isl2 +/-, vs. 0.068±0.005 s for Isl2 KO, p=0.0001) in EMG recordings of Gl muscles. Three animals per group; box plots illustrate distribution with median (center line), first and third quartiles (box boundaries), and 10th-90th percentiles (whiskers); unpaired Student’s t-test; see Supplementary file 2 and source data for detailed n and statistics. ****p<0.0001, ***p<0.001.

Figure 9—source data 1

Measurement of stride width of forelimbs and hindlimbs in footprint analysis of 3-month-old Isl2 +/- and Isl2 knockout (KO) mice for Figure 9c.

Measurement of body and muscle weight of the gluteus and rectus femoris muscles of Isl2 +/- and Isl2 KO mice for Figure 9e.

https://cdn.elifesciences.org/articles/84596/elife-84596-fig9-data1-v2.xlsx
Figure 9—source data 2

Quantification of the average frequency, number of single-motor-unit potentials, and burst activity duration in electromyographic (EMG) recordings of the gluteus (Gl) muscles for Figure 9h–j.

https://cdn.elifesciences.org/articles/84596/elife-84596-fig9-data2-v2.xlsx
Video 2
Example movie of a P14 littermate control Isl2 +/- mouse and an Isl2 knockout (KO) mouse during walking.
Video 3
Example movie of an adult littermate Isl2 +/- and an Isl2 knockout (KO) mouse during walking.

Discussion

The acquisition of specialized motor pools and the specificity of target muscle connections with precision are key steps for effective motor control. We demonstrated that Isl2 is critical for the clustering of motor pool subsets and the formation of dendritic arborization and NMJs in the hindlimb proximal muscles by demonstrating that Isl2-deficient mice display muscle rigidity and abnormal limb coordination. We discuss our findings in the context of genetic programs of Isl2 that direct these connections in a motor pool-specific manner.

Unraveling motor pool-specific roles of Isl2 and the Gdnf-Etv4 pathway in MN development

Our careful observation of Isl2 mutant spinal cords revealed that the clustering of MNs was disorganized, with scattered MMC neurons at all segmental levels, and some LMCl motor pools at the hindlimb level. A previous report and also ours showed that development of the PGC population was compromised (Thaler et al., 2004). Determining the extent to which altered positions of individual motor neuronal populations contribute to the motor function of Isl2 mutant mice remains challenging. Nevertheless, our focused investigation on Etv4-expressing motor pools, particularly Gl and Tfl ones, revealed a significant correlation between mislocalized cell bodies, erroneous connectivity, and abnormal hindlimb movements. Previous studies have addressed the importance of motor neuronal cell body position, especially in genetic contexts related to Hox pathway, such as Foxp1 and Pbx, which establish neuronal subtype differentiation, organization, and connectivity (Dasen et al., 2008; Hanley et al., 2016; Sürmeli et al., 2011). In the absence of these genes, MNs intermingle, leading to a variety of developmental defects in sensorimotor circuits and behaviors. Remarkably, expression of representative motor pool markers, including Etv4, was downregulated in Foxp1-null mice and was spared in Pbx-null mice, implying the involvement of a complex and hierarchical transcriptional network in constructing sensorimotor connectivity in specific motor pools. Herein, we demonstrated that Isl2 plays a crucial role in the correct position and proper function of Etv4-expressing LMC motor pools, particularly at the hindlimb level.

Additional compelling evidence that Isl2 mainly functions in Etv4+ motor pools comes from the striking similarities and differences observed in the mutant phenotypes of Isl2 and the Gdnf-Etv4 pathway, requiring a detailed comparison. Gdnf has been recognized as a critical factor for the survival and innervation of MNs during early MN development (Henderson et al., 1994). Deletion of Gdnf or its receptor Ret or Gfra1 leads to impaired innervation of the peroneal nerve, resulting in no or disrupted axon innervation in major hindlimb muscles, such as Edl, Atib, psoas major, Gmax, and Tfl, to varying degrees (Bonanomi et al., 2012; Gould et al., 2008; Kramer et al., 2006). In contrast, Isl2 mutant mice showed normal axon projection of the peroneal nerve and relatively normal axon innervation in most hindlimb muscles, unlike motor axons of mice deficient in Gdnf or its receptors. Notably, discrepancies between the roles of Gdnf and Isl2 also emerged in types of motor pools affected. The removal of Gdnf, its receptor Ret or Gfra1, or Etv4 has been mostly characterized in CM and LD motor pools, the Etv4-expressing motor pools at the forelimb level, but their roles in the lumbar motor pools have not been addressed (Arber et al., 2000; Haase et al., 2002; Livet et al., 2002; Vrieseling and Arber, 2006). In our study, we observed a selective loss of Etv4 expression only in the lumbar spinal cord of Isl2 KO mice, specifically in Gl and Tfl motor pools. These motor pools displayed a phenotype almost identical to CM and LD motor pools of Etv4-null mice, including mispositioned cell bodies, impaired dendrite and axon arborization, defective NMJs, and reduced sensorimotor connectivity. In addition, acute KD of ISL2 in the chicken neural tube and in MN-specific Isl2 cKO mice was sufficient to abolish Etv4 transcripts, suggesting that Isl2 positively regulates Etv4 expression in a cell-autonomous manner. Thus, Gdnf likely plays broader roles in guiding most LMC projections and also promoting differentiation of Etv4-expressing motor pools at both limb levels. In contrast, Isl2 plays a precise role in fine-tuning the position of MNs, including MMC and LMCl, and also functions in selected lumbar motor pools at the hindlimb, particularly in establishing sensorimotor connectivity during the late period of development.

Nevertheless, Etv4 expression in the CM motor pool at the brachial level was intact in both Isl2 KO mice and cKO mice, implying that another complex genetic program works in conjunction with Isl2 to contribute to segment-specific regulation. For instance, multiple Hox genes control Gdnf signaling, with Hoxc6 activating Ret and Gfra1 expression in rostral brachial spinal cords, regulating terminal arborization of CM and LD motor pools, while Hoxc8 activates Ret and Gfra3 expression in caudal brachial spinal cords, controlling limb innervation of distal nerves (Catela et al., 2022). Similar regulatory pathways involving Hox/Meis in Ret expression have been suggested at lumbar levels, with misexpression of Hoxc10 or Hoxd10 in chicken embryos resulting in Ret expression arising in the neural tube (Catela et al., 2022). Our bulk RNA-seq analysis identified Hoxc11 as one of the DEGs, indicating a potential compromise of proper lumbar identity in the absence of Isl2. However, the expression of Ret and Gfra1 was not changed in Isl2 KO mice when assessed by bulk RNA-seq analysis and in situ hybridization, and overall hindlimb motor projections remained intact. Although further investigation is necessary to fully elucidate the hierarchy among Hox, Gdnf, and Isl2, it is evident that each factor plays a crucial role, to varying degrees, in the proper expression of Etv4 within proximal hindlimb motor pools.

Thaler et al. initially reported that Isl2 KO mice died within a day after birth due to respiratory failure as well as MN developmental defects (Thaler et al., 2004). In contrast, we found that the identical mouse line exhibited a greater survival rate, about 10%. On the other hand, MN-specific Isl2 cKO mice showed similar lethality, with all pups dying on the first day after birth. There are several possibilities to explain the variable mortality among different mouse lines lacking Isl2. First, acute elimination of Isl2 gene in developing MNs could be more detrimental, minimizing potential developmental compensation through altered expression of other genes reconciling the impact of the target KO. Second, the genetic background of the mice may play a role, as it has been reported that different genetic backgrounds can cause varying phenotypes and lethality in KO mouse models (El-Brolosy and Stainier, 2017). For instance, the lethality of Lhx4 embryos was found to be different in C57BL/6J relative to the mixed background (Gergics et al., 2015). Similarly, the genetic background of original Isl2 global null mice could be different from ours, with these mice being occasionally backcrossed onto the C57BL/6 genetic background. Despite the variable survival rates, the defects found in Isl2-null and conditional KO embryos were essentially identical, including scattered lumbar motor pools, absence of Etv4 expression, and defective NMJs, all of which are closely related to the adult behavioral phenotypes found in global Isl2 KO mice. These findings further highlight the critical role of Isl2 in MN development and locomotion. Overall, our study provides a comprehensive analysis of the broad spectrum of phenotypes resulting from removal of Isl2 expression.

Transcriptomic analysis to uncover Isl2-dependent genes for motor pool formation

Recent genome-wide transcriptomic studies have shed light on dynamic transcriptional changes occurring during the development of central nervous system, including spinal MNs (Amin et al., 2021; Delile et al., 2019; Liau et al., 2021; Wang et al., 2022). However, the molecular signature responsible for the organization of multiple individual motor pools has remained elusive. In this study, we investigated the detailed spatiotemporal expression of Isl2 among multiple lineages of MNs using a published scRNA-seq dataset. We successfully segregated the clades of LMC motor pools into six subclusters that encompassed most of the major motor pools at the rostral lumbar segment. Our investigation revealed that sb.LMCl.2 exhibits a molecular signature corresponding to Etv4-expressing Gl, Tfl, and Rf motor pools, with Isl2 expression relatively enriched in this subcluster. Similar LMC subclusters expressing Etv4 was also identified by other groups sharing similar molecular codes (Liau et al., 2023; Wang et al., 2022).

In the current study, we aimed to integrate bulk and scRNAseq analyses datasets to narrow down the motor pools whose terminal differentiation is primarily dependent on the presence of Isl2, as well as to identify potential target genes under the control of Isl2. From the Isl2-deficient lumbar spinal cords, we identified 96 DEGs downregulated, and 87 of them were lumbar-specific. In particular, seven genes (C1ql3, Etv4, A830018L16Rik, Barx2, Kcnab1, Hrk, and Prph) were enriched genes in subcluster sb.LMCl.2, and five of them (C1ql3, Etv4, A830018L16Rik, Barx2, and Kcnab1) belonged to the top 30 of them (see Figure 6—figure supplement 1i, Supplementary file 3b-c). Upon comparing our DEGs with another scRNA-seq dataset analyzed by Liau et al., 2023 we found similar results. Ten genes (Anxa2, A730046J19Rik, Chodl, C1ql3, Etv4, Kcnab1, Lix1, Megf11, Prph, and Uts2b) were also found to be part of the molecular features characterizing the LMCl subcluster called c/rl4 (Lhx1+ Etv4+), confirming the validity of our analyses (Liau et al., 2023). We next aimed to determine whether our approach allowed us to discover new marker genes for designated motor pools. Among the list of enriched genes in sb.LMCl.2 retrieved from scRNA-seq analysis, we demonstrated that three genes Sema5a, Epha3, and Fgf10, exhibited selective expression in Gl motor pools and reduced expression in Isl2null mice according to our in situ hybridization results. Overall, our comprehensive approach, combining bulk and scRNA-seq analyses datasets, has proved highly valuable in identifying new potential motor pool markers and unraveling the genetic program underlying their development.

Our transcriptomic analysis and in vivo validation of candidate targets revealed that numerous genes involved in motor neuronal differentiation and diseases were indeed under the control of Isl2. Among the 96 DEGs downregulated in Isl2 mutant spinal cords, 32 genes (i.e., Adra2c, Arhgap25, C1ql3, C3, Cck, Chrm5, Dkk1, Etv4, Fstl3, Gabra5, Gpr149, Grin2a, Hrk, Htr4, Kcnab1, Kcnj12, Mcf2, Mchr1, Mctp1, Nos1, Ntf3, Oprd1, P2rx4, Pcp4, Pnmt, Prph, Pvalb, Scnn1a, Sema3e, Tmem100, Tmsb15b2, and Tnc) were associated with axon- or synapse-related GO terms, or their protein expression was found in axons or synapses (Dhaese et al., 2009; Fink et al., 2017; Martinelli et al., 2016; Molinard-Chenu et al., 2020; Wu et al., 2019). In addition, 14 genes (i.e., Anxa2, Bcl2l15, C3, Chodl, Fam135b, Hrk, Lix1, Lrp1b, Moap1, Pcp4, Prph, Rnase4, Scg5, and Scnn1a) were reported as ALS- or MN disease-associated genes in the literature (Benoit et al., 2013; Ghavami et al., 2014; Gros-Louis et al., 2004; Lederer et al., 2007; Li et al., 2013; Orr et al., 2020; Sasongko et al., 2010; Sheila et al., 2019; Shu et al., 2022; Sleigh et al., 2014; Su et al., 2022; Wei et al., 2011; Wu et al., 2019). Therefore, approximately 40% of Isl2-dependent DEGs were involved in the maturation and pathological processes of MNs. Prph, a neuronal intermediate filament protein implicated in ALS, showed the largest fold changes (<−0.5 log2-fold change) in our transcriptomic analysis and was selectively downregulated in specific motor pools in Isl2 mutant mice. In addition, earlier extinction of Prph in developing MNs correlated with progressive degeneration of spinal MNs and NMJs, which indicated that the defects observed in Isl2 mutant mice are relevant to early-onset progressive MN disorders such as SMA and ALS (Amin et al., 2015; Amin et al., 2021; Beaulieu and Julien, 2003; Chen et al., 2013; Gros-Louis et al., 2004; Sabbatini et al., 2021). C1ql3 has been shown to promote synapse formation and maintenance in certain excitatory neurons, and its absence results in fewer excitatory synapses with diverse behavioral defects (Martinelli et al., 2016). Given that these MN disease-related genes were all selectively downregulated in specific motor pools that showed deteriorating features in the absence of Isl2, it is likely that proper regulation of these genes by Isl2 ensures proper MN development, whereas dysregulation leads to MN degeneration.

In summary, our findings suggest that pool-specific Isl2 activity is important for the fidelity of motor pool organization and motor circuit connectivity for hindlimb locomotion. Furthermore, Isl2 is responsible for the transcriptional control of a variety of genes involved in MN differentiation, axon development, and synaptic organization, and its absence may give rise to a pathological condition of MNs.

Materials and methods

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Strain, strain background (Mouse, female)C57BL/6JDamul Science Co.
Genetic reagent (Mus musculus)Isl2-/-PMID:14766174MGI:3046260
Genetic reagent (M. musculus)Isl2flox/floxThis paperMGI:109156See Materials and methods, and Figure 3—figure supplement 1a
Genetic reagent (M. musculus)Tg(Mnx1-GFP)1SlpPMID:15201216MGI:3767834
Genetic reagent (M. musculus)B6.Cg-Tg(Thy1-YFP)16Jrs/JJackson Laboratory
PMID:11086982
RRID:IMSR_JAX:003709
Genetic reagent (M. musculus)Olig2tm1(cre)TmjPMID:18046410MGI:3774124
Cell line (human)293[HEK-293]Korean Cell Line BankKCLB# 21573Verified using STR profiling and confirmed to be mycoplasma-free
AntibodyAnti-Olig2 (guinea pig polyclonal)Jessell lab(1:8000)
AntibodyAnti-Hb9 (rabbit polyclonal)PMID:10471502(1:8000)
AntibodyAnti-Hb9 (guinea pig polyclonal)PMID:10482235(1:2000)
AntibodyAnti-GFP (mouse monoclonal)SigmaCat# G6539(1:2000)
AntibodyAnti-Isl1/2 (rabbit polyclonal)PMID:7528105
PMID:8565076
(1:5000)
AntibodyAnti-Isl2 (mouse monoclonal)Santa CruzCat# sc-390746(1:500)
AntibodyAnti-Lhx3 (guinea pig polyclonal)Pfaff lab(1:4000)
AntibodyAnti-ChAT (goat polyclonal)ChemiconCat# AB144P(1:100)
AntibodyAnti-Foxp1 (rabbit polyclonal)AbcamCat# ab16645(1:5000)
AntibodyAnti-Nkx6.1 (goat polyclonal)R&D SystemsCat# AF5857(1:1000)
AntibodyAnti-Isl1 (goat polyclonal)R&D SystemsCat# AF1837(1:1000)
AntibodyAnti-Isl2 (guinea pig polyclonal)PMID:14766174Cat# AF1837(1:8000)
AntibodyAnti-Lhx1 (rabbit polyclonal)AbcamCat# ab14554(1:500)
AntibodyAnti-Scip (guinea pig polyclonal)Dasen lab(1:8000)
AntibodyAnti-Etv4 (rabbit polyclonal)Dasen lab(1:16,000)
AntibodyAnti-Tetramethylrhodamine (rabbit polyclonal)InvitrogenCat# A-6397(1:1000)
AntibodyAnti-vGluT1 (guinea pig polyclonal)SigmaCat# AB5905(1:32,000)
AntibodyAnti-nNOS (rabbit polyclonal)DiaSorin(1:16,000)
AntibodyAnti-pSMAD (rabbit monoclonal)Cell Signaling TechnologyCat# 9516S(1:500)
AntibodyAnti-GFP (rabbit polyclonal)AbcamCat# ab290(1:1000)
AntibodyAnti-Tubulin alpha (rat monoclonal)AbD SerotecCat# MCA77G(1:10,000)
AntibodyAnti-HA (mouse monoclonal)CovanceCat# MMS-101R(1:5000)
AntibodyAnti-Isl2 (mouse monoclonal)DSHBCat# 51.4H9(1:200)
AntibodyAnti-MNR2 (mouse monoclonal)DSHBCat# 81.5C10(1:500)
AntibodyAnti-Digoxigenin-AP, Fab fragments (sheep polyclonal)RocheCat# 11093274910(1:5000)
Sequence-based reagentckISL2 siRNA-461This paperSee Materials and methods, and Figure 5c, Figure 5—figure supplement 2a–b
Sequence-based reagentckISL2 siRNA-605This paperSee Materials and methods, and Figure 5c, Figure 5—figure supplement 2a–b
Peptide, recombinant proteinSP6 RNA polymeraseRocheCat# 10810274001
Peptide, recombinant proteinT7 RNA polymeraseRocheCat# 10881775001
Peptide, recombinant proteinProtector RNase InhibitorRocheCat# 3335399001
Peptide, recombinant proteinRQ1 RNase-Free DNasePromegaCat# M6101
Peptide, recombinant proteinProteinase KVWRCat# E195-5ML
Commercial assay or kitNucleoSpin RNA XSMACHEREY-NAGELCat# 740902.50
Commercial assay or kitpGEM-T Easy Vector SystemsPromegaCat# A1360
Commercial assay or kitDiaStar OneStep RT-PCR kitSolgentCat# DR61-K050
Commercial assay or kitDIG RNA Labeling MixRocheCat# 11277073910
Chemical compound, drugDextran, Tetramethylrhodamine, 3000 MWInvitrogenCat# D3308
Chemical compound, drugAlpha-Bungarotoxin ConjugatesInvitrogenCat# B35451
Chemical compound, drugNBT/BCIP stock solutionRocheCat# 11681451001
Software, algorithmZenZEISSZen Black 2.3 SP1https://www.zeiss.com/microscopy/us/products/microscope-software/zen.html
Software, algorithmFV31S-SWOlympus2.5.1.228https://www.olympus-lifescience.com/
Software, algorithmGraphPad PrismGraphPad softwareVersion 9https://www.graphpad.com/
Software, algorithmImageJNational Institutes of Health1.53k
Software, algorithmR v3.6.0R-projecthttps://www.r-project.org/
Software, algorithmR v4.0.2R-projecthttps://www.r-project.org/
Software, algorithmSeurat v3.2.3R-projectPMID:29608179https://satijalab.org/seurat/get_started.html
Software, algorithmMetascapePMID:30944313https://metascape.org/
Software, algorithmFlaskihttps://flaski.age.mpg.de.

Mice

The Isl2-null, Hb9::GFP and Thy1::YFP mice used in this study have been previously described (Jackson laboratory) (Feng et al., 2000; Thaler et al., 2004). Wild-type C56BL/6 mice (6–8 weeks of age) were purchased from Damul Science Co. (Daejon, Korea). To selectively delete Isl2 from MNs, we generated a novel Isl2 flox mouse (Cyagen, China). The flox strategy was designed by Cyagen using the following procedure. The Isl2 gene (NCBI reference sequence: NM_027397; Ensemble: ENSMUSG00000032318) was located on chromosome 9 in mice. Exons 2–4 were selected as cKO regions. In the targeting vector, the Neo cassette was flanked by SDA sites, and DTA was used for negative selection. C57BL/6N ES cells were used for gene targeting. The KO allele was obtained after specific Cre-mediated recombination. Genotyping assays were designed by Cyagen. First-generation mice were heterozygous for Isl2 flox expression. Heterozygous mice were bred to generate mice with homozygous Isl2 flox expression. The Isl2 flox mouse line was maintained as a homozygous line for breeding with Cre strains. For the Cre line, an MN-specific Olig2Cre line was used (Dessaud et al., 2007; Kong et al., 2015). To obtain Isl2F/KO; Olig2Cre mice, male Isl2 +/-; Olig2Cre mice were crossed with female Isl2F/F mice. The Isl2F/KO or Isl2F/+; Olig2Cre mice from the same litters were used as normal controls. For genotyping PCR, genomic DNA was isolated from mouse ear punches digested in DirectPCR (tail) (VIAGEN) and Proteinase K (New England Biolabs). The following primers were used for genotyping PCR: 5’-TGG GAC TAC GGG GTT GTA CTT-3’ and 5’-GTT CTG GAG AGC AAG TTG GGA AT-3’ to detect a wild-type allele (274 bp) and a flox allele (410 bp). All experiments used protocols approved by the Animal Care and Ethics Committees of the Gwangju Institute of Science and Technology (GIST). The day when a vaginal plug was detected was designated as embryonic day 0.5 (E0.5).

Immunohistochemistry and in situ hybridization

Request a detailed protocol

Immunohistochemistry and in situ hybridization were performed as described previously (Song et al., 2009). The following antibodies were used: guinea pig anti-Olig2 (Wichterle et al., 2002), rabbit and guinea pig anti-HB9 (Harrison et al., 1999), rabbit anti-GFP (Invitrogen), mouse anti-GFP (Sigma), rabbit anti-Isl1/2 (Pfaff et al., 1996), mouse anti-Isl2 (Santa Cruz), guinea pig anti-Lhx3 (Thaler et al., 2004), goat anti-ChAT (Chemicon), rabbit anti-Foxp1 (Abcam), goat anti-Nkx6.1 (R&D Systems), goat anti-Isl1 (R&D Systems), guinea pig anti-Isl2 (Thaler et al., 2004), rabbit anti-Lhx1 (Abcam), guinea pig anti-Scip (Dasen et al., 2005), rabbit anti-Etv4 (Dasen et al., 2005), rabbit anti-Tetramethylrhodamine (Invitrogen), guinea pig anti-vGluT1 (Millipore), rabbit anti-nNOS (Diasorin), and rabbit anti-pSMAD (Cell Signaling Technology). For NMJ staining, muscles were harvested and were prepared for immunostaining as whole-mount samples. Samples were immunostained with rabbit anti-GFP (Abcam) and Alexa Fluor 555 α-BTX (Invitrogen). For in situ hybridization, embryonic mouse cDNA at E12.5 was used to generate riboprobes for Isl2, Ret, Gfra1, A730046J19Rik, Anxa2, Kcnab1, Prph, C1ql3, Epha3, Fgf10, and Sema5a. HH stage 25 chicken embryonic cDNA was used to generate riboprobes for chicken ALDH1A2 and ETV4 using a one-step RT-PCR kit (Solgent).

Western blot analysis

Request a detailed protocol

293[HEK-293] cells (KCLB No. 21573) were purchased from Korean Cell Line Bank. The cell lines were verified using STR profiling and confirmed to be mycoplasma-free. 293[HEK-293] cells were cultured and transfected with an expression plasmid containing N-terminal truncated chick ISL2 (aa 38–356) with an HA tag and ISL2 siRNA using Lipofectamine (Invitrogen). After 36 hr, the cells were dissociated and lysed in lysis buffer for western blotting. The western blotting procedure followed previous protocols (Song et al., 2009). Rat anti‐α‐tubulin (AbD Serotec) and mouse anti-HA (Covance) antibodies were used.

Chick in ovo electroporation

Request a detailed protocol

In ovo electroporation was performed as described previously (Kim et al., 2016). A DNA solution was injected into the lumen of the spinal cord of 10–12 chicken embryos at the HH stage, and the embryos were harvested at HH stage 25. To knock down ISL2, ISL2 siRNAs (siRNA-461 sense 5’-GGA CGG UGC UGA ACG AGA A-3’, siRNA-461 antisense 5’-UUC UCG UUC AGC ACC GUC C-3’; siRNA-605 sense 5’-GCU GCA AGG ACA AGA A-3’, siRNA-605 antisense 5’-UUC UUC UUG UCC UUG CAG C-3’) were electroporated. Harvested chicken embryos were processed for immunohistochemistry and in situ hybridization. Rabbit anti-GFP (Invitrogen), mouse anti-Isl2 (DSHB), and mouse anti-MNR2 (DSHB) antibodies were used. Riboprobes for chicken ALDH1A2 and ETV4 were amplified from total RNA extracted from HH 25 chicken embryos using a one-step RT-PCR kit (Solgent). For image analysis, consecutive sections at the lumbar level were collected from each embryo. In each ventral quadrant, the number of MNs was counted, and the signal intensity was measured using the ImageJ software (NIH). The intensity values of the electroporated side were normalized to the value of the nonelectroporated side.

Intramuscular injections of tracers

Request a detailed protocol

To label MN dendrites, the target muscles were injected with Rh-Dex (3000 MW, Invitrogen). In brief, P4 mice were anesthetized on ice, and the muscles were exposed by making a small incision in the skin. Approximately 2 μl of 10% Rh-Dex was injected at a single site, and the skin was sutured. After 3 days, the spinal cords were analyzed for dextran labeling, and the muscles were isolated to check the injection site. Analyses of dendritic arborization patterns of the Gl motor pools primarily focused on the L3 and L4 spinal segmental levels. Free-floating 80-μm-thick transverse sections were immunostained for further analysis.

For sensory bouton analysis, P16 mice were anesthetized, and the skin was incised to expose the muscle of interest. Muscles were injected at multiple sites with 10% tetramethylrhodamine dextran (3000 MW, Invitrogen) using a pulled glass microelectrode. At P21, the spinal cords were harvested and processed for immunostaining.

EMG analysis

Request a detailed protocol

For EMG analysis, we used multistranded, Teflon-coated, annealed stainless steel wire (A-M systems, Cat. No. 793200) electrodes that were soldered to an IC socket. After mice were anesthetized with avertin, we implanted the electrodes subcutaneously into the middle of the left and right Gl muscles and secured the connector to the head. The mice were allowed to recover in their cages for at least 2 days before recording began. After the recovery period, we placed the animals in an open field where they could move freely. During locomotor activity, EMG signals were digitized and stored using Clampex 9.2 data acquisition software, while video recordings were made to monitor walking movement for 30 min. EMG and video data were recorded simultaneously. The EMG data was bandpass filtered at 200 Hz to 1 kHz using Clampfit 10.3. We measured and analyzed the number, frequency, and duration of EMG peaks over 1 s.

Tail suspension test and gait analysis

Request a detailed protocol

For the tail suspension test, we suspended each mouse above the cage by its tail and captured its posture on a camera. In the gait analysis, we used 3-month-old Isl2 KO and littermate female controls. We applied nontoxic water-based paints (red on forepaws and black on hindpaws) to the mouse paws and allowed the mice to walk on white paper. We performed three consecutive trials and measured the stride widths of the forelimbs and hindlimbs.

RT-PCR

Request a detailed protocol

E12.5 embryonic spinal cords were harvested in ice-cold PBS and total RNA was extracted using NucleoSpin RNA XS (MACHEREY-NAGEL). We used 10 ng of total RNA per 15 μl reaction for the RT-PCR with the following primers: (Gapdh, 5’-GGA GAA ACC TGC CAA GTA TGA-3’, 5’-CCT GTT GCT GTA GCC GTA TT-3’, Etv4 5’-GGT GAT GGA GTG ATG GGT TAT G-3’, 5’GCC TGT CCA AGC AAT GAA ATG-3’). The PCR products were amplified using a one-step RT-PCR kit (Solgent). Band densitometry was quantified for Etv4 and Gapdh using the ImageJ software.

Image acquisition

Request a detailed protocol

For data analysis, images were captured using epifluorescent LSM780 (Zeiss) and FV3000RS (Olympus) confocal microscopes using the Axiovision, ZEN software (Zeiss) and the FV31S software (Olympus).

Radial plot

Request a detailed protocol

To analyze dendritic radiality, motor pools were subdivided into eight divisions, with the center aligned with the center of the motor pool. For each octant, the mean pixel intensity of dendrites, excluding the cell bodies, was quantitated using the ImageJ software and normalized to the octant with the highest value.

Contour and density plots

Request a detailed protocol

MN soma coordinates were acquired with respect to the midline using ImageJ software. To assign x and y coordinates, we normalized experimental variations from different spinal cord sizes, shapes, and sections. Contour, spatial, and density plots were created using the ‘ggplot2’ package in R-4.0.5. Contour plots were calculated using ‘MASS::kde2d()’ by performing a 2D kernel density estimation and displaying the results with contours for the distribution of neuron positions.

Quantification of sensory boutons

Request a detailed protocol

For quantitative analysis of vGluT1+ sensory boutons in Rh-Dex+ Gl motor pools, spinal cords were sectioned at a thickness of 40 μm. A z-series of 0.74 μm optically scanned confocal images using a 30× objective were acquired for quantitative analysis. The number of vGluT1+ puncta within each somatic compartment was quantified across the Rh-Dex+ MNs. vGluT1 intensity was determined using ImageJ software, and vGluT1+ synapses were counted manually (Wang et al., 2019).

Quantification of NMJ morphology and imaging

Request a detailed protocol

NMJ images were acquired on LSM780 (Zeiss) and FV3000RS (Olympus) confocal microscopes. To analyze NMJs, maximum intensity projections were created using ZEN (Zeiss) and FV31S (Olympus) software. Low-magnification images were acquired using 10× and 20× objectives. For NMJ morphology assessment, confocal settings were optimized: 640×640 frame size, 30× objective, ×6.0 zoom, and 0.74 μm z-stack interval for quantification in P14 muscles; and 800×800 frame size, 30× objective, ×4.5 zoom, and 0.74 μm z-stack interval for quantification in P28 muscles. For quantification of Gl muscles, the number of NMJs was quantified along the entire Gl nerve within the Gl muscles in P0 and P14. Quantification of the length of secondary branches, end plate area occupied by motor axons or AChR clusters in P0, and the nerve terminal area, AChR cluster area, compactness, and fragmentation in P14 and P28 was performed using ImageJ software (Jones et al., 2016). The number of axons extending the border of the target muscles was calculated by quantifying the overshooting axons per 100 μm of end plate. The proportion of polyinnervated, faint, denervated NMJs was quantified within three to four random fields per animal. The proportion of polyinnervated NMJs was quantified in Gl nerves adjacent to Tfl muscles. The number of axonal inputs was quantified in individual NMJs in P14. In Tfl muscles, we quantified the number of NMJs along the entire nerve within the Tfl muscles at P14, and the number of axonal inputs in individual NMJs at P28.

Skeletal staining and X-ray imaging

Request a detailed protocol

E17.5 mouse embryos were prepared by removing the skin, viscera, and adipose tissues and fixed in 95% ethanol for at least 5 days. The tissues were incubated in acetone for 2 days to remove fat and stained in staining solution (1 volume of 0.3% Alcian blue in 70% ethanol, 1 volume of Alizarin red in 95% ethanol, 1 volume of glacial acetic acid, and 17 volumes of 70% ethanol) for 3 days at 37°C. Stained tissues were rinsed in distilled water and cleared in 1% KOH, 20% glycerol in 1% KOH, 50% glycerol in 1% KOH, and 80% glycerol in 1% KOH, sequentially, for several days each. Cleared embryos were stored in 100% glycerol before imaging.

Whole-mount immunostaining for axon tracing

Request a detailed protocol

E11.5, E12.5, and E13.5 Hb9::GFP embryos were dissected, and samples were immersed in 4% PFA and washed in PBS overnight. Tissues were cleared using a three-dimensional imaging kit (Binaree). Images were captured using a Zeiss confocal microscope using the ZEN software (Zeiss).

RNA-seq and bioinformatics analysis

Request a detailed protocol

For the RNA-seq analysis of ventral brachial and lumbar spinal cords, E12.5 embryonic spinal cords (three heterozygote animals and three homozygote animals) were harvested and trimmed around the Hb9::GFP reporter in ice-cold RNase-free PBS. Total RNA was extracted using NucleoSpin RNA XS, Micro kit (MACHEREY-NAGEL) for RNA purification, and RNA quality and quantity were assessed using a 2100 Bioanalyzer System (Agilent). All samples showed high-quality scores between 9.6 and 10 RIN. 500 ng of total RNA per sample was used for the construction of sequencing libraries, which was amplified using the TruSeq Stranded Total RNA LT Sample Prep Kit (Gold). Library preparation and bulk RNA-seq were performed by Macrogen (Seoul, Korea) using the manufacturer’s reagents and protocols. Libraries were sequenced on an illumine NovaSeq platform. Raw sequencing files were assessed by FastQC (v0.11.7) for quality, and sequence reads were trimmed for adaptor sequence and low-quality sequence by Trimmomatic (v0.38). The processed reads were mapped to the mm10 mouse reference genome using HISAT2 (v2.1.0) and Bowtie2 (v2.3.4.1) for further processing. Assembly of aligned reads and abundance estimations were performed using StringTie (v2.1.3b), and fragments per kilobase of transcript per million fragments mapped values and transcript per million values were acquired. Analysis of DEGs was conducted using DESeq2.

GO analysis

Request a detailed protocol

GO analysis based on bulk RNA-seq data was performed using Metascape (http://metascape.org) (Zhou et al., 2019).

Statistical representation in figures

Request a detailed protocol

Error bars represent SEM. Box plots show 25th percentile, median, and 75th percentile values, with whiskers indicating 10th and 90th percentile values. Outlying values are shown as symbols. Statistical analyses were performed with GraphPad Prism 9 (GraphPad Software, San Diego, CA, USA).

Use of published datasets

Request a detailed protocol

For the evaluation and comparison of candidate genes downregulated in Isl2-deleted embryos, we compared our gene lists with a previously published dataset (ArrayExpress accession: E-MTAB-10571) (Amin et al., 2021). The raw data from the study by Amin et al., 2021. were obtained from ArrayExpress accession: E-MTAB-10571. Reads from wild-type samples (E12 Hb9::GFP) were mapped to the mm10 mouse reference genome, and raw unique molecular identifiers were produced using Cell Ranger (version 3.1.0, 10x Genomics). The expression matrices were processed in R (v.3.6.0) using Seurat (v.3.2.3). Cells with fewer than 1000 detected transcripts were excluded and cells with over 8% of mitochondrial gene read counts were excluded. Filtered gene-barcode matrices were normalized using the NormalizeData function with LogNormalize, and the top 2000 variable genes were identified using the ‘vst’ method in FindVariableFeatures. Gene expression matrices were scaled using the ScaleData function. Principal component analysis (PCA) and UMAP were performed using the 20 principal components. To remove the batch effect, RunFastMNN was applied to the PCA matrix with SampleID as the batch key. Dimensionality reduction was performed separately for all cells and the interneuron-excluded subset. Cells were partitioned into 31 clusters with a resolution parameter of 2. Neuron subtype identification was performed as previously described (Amin et al., 2021). MN subtypes were identified according to the expression of known marker genes. The number of cells in each neuron subtype was as follows: 19 cells in pMNs, 672 cells in immature MNs, 3464 cells in postmitotic MNs, and 1135 cells in interneurons.

Further downstream analyses within LMC clusters were performed using Seurat v.4.0.2 R packages. To analyze motor pools at the L1 to L4 levels, we isolated LMC subclusters that expressed Hoxd9, Hoxc10, Hoxd10, Hoxa11, Hoxc11, and Hoxd11. As a result, LMC clusters were further divided into seven clades composed of 276 cells: sb.LMCl.1.v, sb.LMCl.1.ta, sb.LMCl.2, sb.LMCm.1, sb.LMCm.2, and sb.LMCm.3. Lhx1, an LMCl marker, was enriched in sb.LMCl.1.v, sbLMCl.1.ta, and sb.LMCl.2 clades, whereas Isl1, an LMCm marker, was enriched in LMCm.1, LMCm.2, and LMCm.3. Visualized data were plotted using VinPlot, DotPlot, DoHeatmap, and FeaturePlot functions from the Seurat R packages.

Data availability

Sequencing data have been deposited in GEO under accession codes GSE217297.

The following data sets were generated
    1. Song MR
    2. Lee Y
    (2022) NCBI Gene Expression Omnibus
    ID GSE217297. Transcriptional control of motor pool formation and motor circuit connectivity by the LIM-HD protein Isl2.

References

    1. Vanderhorst VG
    2. Holstege G
    (1997)
    Organization of lumbosacral motoneuronal cell groups innervating hindlimb, pelvic floor, and axial muscles in the cat
    The Journal of Comparative Neurology 382:46–76.

Decision letter

  1. Paschalis Kratsios
    Reviewing Editor; University of Chicago, United States
  2. Marianne E Bronner
    Senior Editor; California Institute of Technology, United States
  3. Paschalis Kratsios
    Reviewer; University of Chicago, United States
  4. Niccolò Zampieri
    Reviewer; Max Delbrück Center for Molecular Medicine, Germany

Our editorial process produces two outputs: (i) public reviews designed to be posted alongside the preprint for the benefit of readers; (ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "Transcriptional control of motor pool formation and motor circuit connectivity by the LIM-HD protein Isl2" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Marianne Bronner as the Senior Editor. The following individuals involved in the review of your submission have agreed to reveal their identity: Paschalis Kratsios (Reviewer #1); Niccolò Zampieri (Reviewer #2).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

1) The issue of cell autonomy was raised by all three reviewers because the study primarily focuses on global Isl2 KO mice. Additional characterization of the conditional Olig2Cre;Isl2 mice would significantly strengthen the conclusions. The authors do not need to repeat every experiment in the conditional Olig2Cre;Isl2 mice, but the reviewers strongly encourage the authors to conduct the same behavioral and connectivity experiments in the conditional mice.

2) The writing is at times not very clear and makes it difficult to appreciate the findings and their relevance in the context of the literature. Please address specific comments by reviewer #1 (presentation of data on Figure 2; grammar issues and typos) and reviewer #2 (the Introduction is difficult to read; more emphasis should be given to what is known about Isl2 and Pea3 mutant mice, which is very relevant for the paper).

3) Based on this study and previous studies (e.g., Catela et al. 2016, Cell Reports, PMID: 26904955), it seems that there is differential regulation of Pea3 at brachial and lumbar levels. The Pea3 phenotypes described by Arber et al. were very dramatic-almost complete lack of innervation. Here the phenotypes appear to be less so and mostly at the NMJ, suggesting Pea3 partially controls aspects of the lumbar motor pool properties. So, if Isl2 regulates Pea3 differentially, it's possible that Pea3 is not gdnf-dependent at lumbar levels (is this known?). The explant experiments aren't that convincing because they contain all lumbar motor axons-if only a few pools are regulated by Isl2 why do none of them respond to gdnf. The authors should have a more thoughtful discussion about the gene regulatory networks at brachial vs lumbar levels and what the significance might be. They also must cite Catela et al. 2016, which nicely showed that Ret expression is dependent on brachial Hox genes, and that Hoxc10/d10 can also induce Ret expression at lumbar levels. They should incorporate these findings in their discussion.

4) Improve quantifications: The way MN numbers are given throughout the manuscript is confusing. The data should be recalculated and presented as average MN numbers/mouse (n=1 mouse and not 1 section). In Figure 2, quantification would be really good to have at both stages. The statistical difference in MMC and PGC numbers (Figure 2f) observed between WT and KO mice should be briefly commented on, in light of Thaler's results and in the context of the ectopic motor neurons.

5) Figure 3: The markers used in Figure 3a do not allow to clearly distinguish MMC, LMCm and LMCl neurons. A different combination should be used such as Lhx3 (MMC) Isl1 (LMCm) and Hb9 (LMCl). In addition, spatial analysis of columnar and divisional segregation should be provided to better support the statement.

6) Further experimentation and/or discussion is needed to clarify which exact motor pools are affected in Isl2 mutant mice. See comments by reviewers 1 and 3.

Reviewer #1 (Recommendations for the authors):

Figure 2 is confusing. The presentation and discussion of the data in Results (lines 149-162) need to be improved. The authors should also present their data more clearly in the context of Thaler et al. (2004), as that study found increased numbers of somatic motor neurons and decreased numbers of visceral motor neurons in the thoracic spinal cord. By doing so, the novelty of the current study will be clearer. Perhaps, switching the order of Figures 2 and 3 would address this issue and improve the flow.

In the transcriptomic analysis, the authors do a remarkable job to identify motor neuron-expressed genes and define subclusters of LMC motor pools. Although the RNA-Seq is done in bulk and without motor neuron specificity, the authors leverage published RNA-Seq data to ascribe MN specificity to their dataset. Hence, they identified several putative targets, which they validate with ISH (Figure 6H). However, they are strongly encouraged to also validate with ISH putative markers of the subcluster LMCl.2 in Isl2 global or conditional mice. This experiment would significantly strengthen the authors' conclusions (lines 356-366) about which motor pools are affected by Isl2 loss, especially because Pea3 is a marker for multiple motor pools.

The study will be strengthened by conducting hindlimb locomotion analysis in Olig2Cre;Isl2 fl/fl mice.

The text can be significantly improved by eliminating typos and improving grammar. The authors should go carefully through the text, sentence by sentence, and make the necessary changes.

In the first section of Results, the authors can cite Thaler et al. (2004), as their results are consistent with that study.

Reviewer #2 (Recommendations for the authors):

In my opinion, the writing is at times not very clear and makes it difficult to appreciate the findings and their relevance in the context of the literature. The experiments are well executed and appropriately analyzed, there are a few things that can be improved, more to it in the following detailed comments:

1) I find the introduction difficult to read, it is more like a list of facts than a coherent narrative to properly frame the work. For example, the bit about neuromesodermal progenitors (lines 66-71), while interesting does not fit with the rest of the story. Also, at times the balance between subjects seems a bit off, more emphasis should be given to what is known about Isl2 and Pea3 mutant mice, which is very relevant to the paper.

2) Line 150: it is not specified which Isl2 KO mice is used. At the very least there should be a reference (Thaler et al., 2004). However, here it is very important to comment that these are the same mice that have been used in a previous study characterizing the role of Isl2 in spinal motor neuron development. In addition, the mice were originally reported to "die within 24 hr of birth" (Thaler et al., 2004), but the authors found that "approximately 10% survived" (lines 315-317). This important information should be introduced at the beginning when the mouse line is first introduced. It will be immediately relevant as in line 160 adult isl2 KO mice are used.

3) Line 159: "Mislocalized motor neurons remained in the adult Isl2-null lumbar spinal cord with a similar number of motor neurons" Mislocalised motor neurons are not very evident from pictures provided in figure 2h higher magnifications would be helpful. Regarding the "similar number", I could not find quantification of mislocalised motor neurons either in embryos or adults, thus a quantitative statement is not justified. Indeed, quantification would be really good to have at both stages.

4) The statistical difference in MMC and PGC numbers (Figure 2f) observed between WT and KO mice should be briefly commented on, in light of Thaler's results and in the context of the ectopic motor neurons.

5) Line 169: "We similarly found that MMC neurons were scattered dorsolaterally and located ectopically, either within the LMC columns or immediately dorsal to the motor columns in Isl2-deficient (Figure 3a)" The markers used in Figure 3a do not allow to clearly distinguish MMC, LMCm and LMCl neurons. A different combination should be used such as Lhx3 (MMC) Isl1 (LMCm) and Hb9 (LMCl). In addition, spatial analysis of columnar and divisional segregation should be provided to better support the statement.

6) Figure 4b: I believe there must be a mistake as it is clear from the pictures in 4a that there are very few Pea3+ neurons in Isl2 KO mice, however in the plots in 4b (last row the three plots on the right) there are huge red contours for Pea3 neurons more or less similar to control mice. Indeed, the difference is clearly evident in the dot plots in Figure 4 supp 1b.

7) It is great that the authors generated an Isl2 conditional allele, however, its phenotype in combination with olig2::cre is not described other than in the experiments in Figure 5 aimed at studying the downregulation of Pae3. Does the Isl2 cKO mouse recapitulate the overall defects observed in Isl2 KO in terms of motor neuron generation, spatial organisation, connectivity, and behaviour? Comparison of the two can be very informative especially regarding the behavioural and sensorimotor connectivity phenotypes as many other cells (i.e., all somatosensory neurons) express Isl2 during development.

8) Do you observe arborization, sensorimotor connectivity, and NMJ defects shown for Gl motor neurons also in Rf and Tfl motor pools? It seems like it would be the case but it would be a great addition in order to generalise the role of Isl2-Pea3 transcriptional axis on all pools expressing it.

Reviewer #3 (Recommendations for the authors):

Specific concerns:

1) The behavioral phenotypes seen in the videos appear to be very dramatic, especially in the hindlimbs. However, the authors show that most hindlimb muscles are properly innervated, with the exception of the Gl. For example, the Rf seems to be properly innervated but its weight is dramatically reduced in Figure 9. What is the cause for this? Do the authors believe that the behavior correlates to defects in Gl development and connectivity or are there non-MN contributions to these phenotypes from sensory neurons and interneurons?

2) The connection between Isl2/gdnf signaling and Pea3 expression is not experimentally supported. The authors do show a specific downregulation of Pea3 at lumbar levels of the spinal cord but whether this is mediated through gdnf is not established. Is Ret or any other component of this signaling pathway downregulated in the RNA-seq data? The fact that lumbar explants do not respond to gdnf may suggest this but the connection is not explicitly shown.

3) The loss of synapses in Fig7b is interpreted as Isl2 controlling Gl connectivity. Since this is not a conditional allele, the synaptic loss could be due to defects in sensory neurons. The authors also show an ectopic soma (although this is not really well defined). Is there a difference between ectopic and Gl MNs that settle in the correct position? Are the authors suggesting that cell body position influences connectivity in a way that ectopic MNs lose more connections? This is not quantified or discussed.

4) The way MN numbers are given throughout the manuscript is confusing. The authors quantify MN/section. However, especially for motor pool counts, this results in a lot of variability since at the beginning and end of each motor pool you have 1-2 MNs, while higher numbers are in the middle. The data should be recalculated and presented as average MN numbers/mouse (n=1 mouse and not 1 section). This would reduce variability within the datasets.

5). Figure 1: The authors discuss the different spatiotemporal expressions of Isl1 and Isl2 but this is not shown (maybe inferred from RNAseq). Only one time point is shown in the figure (e12.5). It would be nice to show the temporal expression of both TFs at different embryonic time points and also in combination with motor pool markers. For example, the overlap of Isl2 with Pea3 at brachial and lumbar levels is not shown. Showing which motor pools express high and sustained levels of Isl2 in this figure would be a nice setup for the rest of the story.

6). Figure 2-4: MN numbers should be recalculated and presented so that n=1 mouse. It is likely that this might change some of the data. For example, in Figure 3a, there is a visible reduction of Scip+ MNs. When the source data for Figure 3c was reanalyzed, there seems to be a 30% reduction of FCU neurons with a P=0.03 (not significant but close, perhaps would be significant if n>3). Is Isl2 highly expressed in Scip+ MNs at brachial levels (see point for Figure 1). Also the MN positional changes should be quantified in these figures.

7). Figure 6: After comparing their RNAseq data to scRNAseq the authors mention that only 50% seem to be MN-specific. Is this technical (not deep enough sequencing in the scRNA-seq datasets) or are the rest of the genes in interneurons? This relates to the comment about other affected populations in the null mouse. In the in situ data, the ventrolateral pools are indicated, however, the changes appear to be broader, affecting the entire lumbar spinal cord. The nNos data should be shown separately since that is in the thoracic spinal cord.

8). Figure 7: In the text, it is mentioned that the Pea3 pools are Gl, Rf, and Tfl and that a retrograde tracer was injected into individual muscles. However, only the Gl data are shown. What about the other pools? Do they have normal dendrites? It seems that Pea3 is not the whole story as Rf has normal NMJ innervation. What about sensory innervation? Is that synapse loss specific to Gl MNs? What about Rf and Tfl? What about non-Pea3+ pools?

https://doi.org/10.7554/eLife.84596.sa1

Author response

Essential revisions:

1) The issue of cell autonomy was raised by all three reviewers because the study primarily focuses on global Isl2 KO mice. Additional characterization of the conditional Olig2Cre;Isl2 mice would significantly strengthen the conclusions. The authors do not need to repeat every experiment in the conditional Olig2Cre;Isl2 mice, but the reviewers strongly encourage the authors to conduct the same behavioral and connectivity experiments in the conditional mice.

Initially, our experiments primarily involved the use of global Isl2 KO mice, and more recently, we generated Isl2 conditional mice to address the issue of cell autonomy. Through meticulous observation, we identified that about 10% of global Isl2 KO survived, allowing us to effectively examine their postnatal phenotypes. Regrettably, the Isl2 conditional KO (cKO) mice showed a different survival rate, as they all succumbed within a day after birth. Thus, our ability to perform experiments utilizing postnatal animals was limited.

Detailed results and the interpretation surrounding the observed mortality discrepancy in these two mouse lines have been comprehensively presented in Table S1, Supplementary Videos 1-3, and elaborated within the Results and Discussion section.

Page 10: “A previous study by Thaler et al. (2004) had established that Isl2 KO mice were unable to survive beyond a day after birth (Thaler et al., 2004). Intriguingly, our study revealed that Isl2 cKO mice also died on the first day after birth, while 11% of Isl2 KO mice managed to survive (Supplementary Table S1, Supplementary Video S3). While the precise underlying cause for this disparity in survival rates among Isl2 KO mutant mouse lines requires further investigation, we were able to examine adult motor neurons in the absence of Isl2.”

Page 16: “Thaler et al. initially reported that Isl2 KO mice died within a day after birth due to respiratory failure as well as motor neuron developmental defects (Thaler et al., 2004). In contrast, we found that the identical mouse line exhibited a greater survival rate, about 10%. On the other hand, motor neuron-specific Isl2 conditional KO mice showed similar lethality, with all pups dying on the first day after birth. There are several possibilities to explain the variable mortality among different mouse lines lacking Isl2. First, acute elimination of Isl2 gene in developing motor neurons could be more detrimental, minimizing potential developmental compensation through altered expression of other genes reconciling the impact of the target knockout. Second, the genetic background of the mice may play a role, as it has been reported that different genetic backgrounds can cause varying phenotypes and lethality in knockout mouse models (El-Brolosy and Stainier, 2017). For instance, the lethality of Lhx4 embryos was found to be different in C57BL/6J relative to the mixed background (Gergics, Brinkmeier, and Camper, 2015). Similarly, the genetic background of original Isl2 global null mice could be different from ours, with these mice being occasionally backcrossed onto the C57BL/6 genetic background. Despite the variable survival rates, the defects found in Isl2 null and conditional KO embryos were essentially identical, including scattered lumbar motor pools, absence of Etv4 expression, and defective NMJs, all of which are closely related to the adult phenotypes found in global Isl2 KO mice. These findings further highlight the critical role of Isl2 in motor neuron development and locomotion. Overall, our study provides a comprehensive analysis of the broad spectrum of phenotypes resulting from removal of Isl2 expression.”

Nevertheless, we successfully investigated the progress of NMJ development in Isl2 cKO at E18.5, just prior to birth, and identified congruent developmental anomalies comparable to Isl2 null mice. Consequently, the phenotypes we observed during the embryonic stage were identical in both mouse lines, i.e., dispersed motor neurons, Etv4 expression absence in selective lumbar motor pools, disrupted NMJ formation within those pools. The new obtained data, along with pertinent explanations, have been integrated into Figure 8—figure supplement 2 and incorporated within the Results and Discussion section, as outlined below.

Page 12: “Due to the postnatal lethality of Isl2 cKO, we analyzed the muscles of these mice at E18.5, prior to birth, when NMJ formation is in progress (Figure 8—figure supplement 2). GFP-labeled axons arrived in Isl2-deficient Gl muscles; however, their terminal arbors were defasciculated and misrouted. Fluorescence intensity of α-bungarotoxin (BTX) at the NMJs was greatly reduced in Isl2 cKO. Furthermore, AChR clustering appeared to be incomplete, showing slender and elongated NMJs, or some NMJs were abnormally concentrated in proximity to the primary axon shaft in Isl2 cKO (Figure 8—figure supplement 2b).”

2) The writing is at times not very clear and makes it difficult to appreciate the findings and their relevance in the context of the literature. Please address specific comments by reviewer #1 (presentation of data on Figure 2; grammar issues and typos) and reviewer #2 (the Introduction is difficult to read; more emphasis should be given to what is known about Isl2 and Pea3 mutant mice, which is very relevant for the paper).

We have thoroughly revised the Introduction, Result and Discussion section of our manuscript, aiming to offer a more comprehensive and coherent analysis of our results and their correlation with prior research in the field. Furthermore, the manuscript has undergone rigorous editorial refinement by a professional editor.

3) Based on this study and previous studies (e.g., Catela et al. 2016, Cell Reports, PMID: 26904955), it seems that there is differential regulation of Pea3 at brachial and lumbar levels. The Pea3 phenotypes described by Arber et al. were very dramatic-almost complete lack of innervation. Here the phenotypes appear to be less so and mostly at the NMJ, suggesting Pea3 partially controls aspects of the lumbar motor pool properties. So, if Isl2 regulates Pea3 differentially, it's possible that Pea3 is not gdnf-dependent at lumbar levels (is this known?). The explant experiments aren't that convincing because they contain all lumbar motor axons-if only a few pools are regulated by Isl2 why do none of them respond to gdnf. The authors should have a more thoughtful discussion about the gene regulatory networks at brachial vs lumbar levels and what the significance might be. They also must cite Catela et al. 2016, which nicely showed that Ret expression is dependent on brachial Hox genes, and that Hoxc10/d10 can also induce Ret expression at lumbar levels. They should incorporate these findings in their discussion.

We appreciate the Reviewer’s insightful comments. As suggested, we have cited the work of Catela et al. and improved the text with a more comprehensive description and discussion, as presented below.

Page 15: “Additional compelling evidence that Isl2 mainly functions in Etv4+ motor pools comes from the striking similarities and differences observed in the mutant phenotypes of Isl2 and the GDNF-Etv4 pathway, requiring a detailed comparison. GDNF has been recognized as a critical factor for the survival and innervation of motor neurons during early motor neuron development (Henderson et al., 1994). Deletion of GDNF or its receptor Ret or Gfra1 leads to impaired innervation of the peroneal nerve, resulting in no or disrupted axon innervation in major hindlimb muscles, such as Edl, Atib, psoas major, Gmax, and Tfl, to varying degrees (Bonanomi et al., 2012; Gould, Yonemura, Oppenheim, Ohmori, and Enomoto, 2008; Kramer et al., 2006). In contrast, Isl2 mutant mice showed normal axon projection of the peroneal nerve and relatively normal axon innervation in most hindlimb muscles, unlike motor axons of mice deficient in GDNF or its receptors. Notably, discrepancies between the roles of GDNF and Isl2 also emerged in types of motor pools affected. The removal of GDNF, its receptor Ret or Gfra1, or Etv4 has been mostly characterized in CM and LD motor pools, the Etv4-expressing motor pools at the forelimb level, but their roles in the lumbar motor pools have not been addressed (Arber, Ladle, Lin, Frank, and Jessell, 2000; Haase et al., 2002; Livet et al., 2002; Vrieseling and Arber, 2006). In our study, we observed a selective loss of Etv4 expression only in the lumbar spinal cord of Isl2 KO mice, specifically in Gl and Tfl motor pools. These motor pools displayed a phenotype almost identical to CM and LD motor pools of Etv4-null mice, including mispositioned cell bodies, impaired dendrite and axon arborization, defective NMJs, and reduced sensorimotor connectivity. In addition, acute knockdown of Isl2 in the chicken neural tube and in motor neuron-specific Isl2 cKO mice was sufficient to abolish Etv4 transcripts, suggesting that Isl2 positively regulates Etv4 expression in a cell-autonomous manner. Thus, GDNF likely plays broader roles in guiding most LMC projections and also promoting differentiation of Etv4-expressing motor pools at both limb levels. In contrast, Isl2 plays a precise role in fine-tuning the position of motor neurons, including MMC and LMCl, and also functions in selected lumbar motor pools at the hindlimb, particularly in establishing sensorimotor connectivity during the late period of development.

Nevertheless, Etv4 expression in the CM motor pool at the brachial level was intact in both Isl2 KO mice and cKO mice, implying that another complex genetic program works in conjunction with Isl2 to contribute to segment-specific regulation. For instance, multiple Hox genes control GDNF signaling, with Hoxc6 activating Ret and Gfra1 expression in rostral brachial spinal cords, regulating terminal arborization of CM and LD motor pools, while Hoxc8 activates Ret and Gfra3 expression in caudal brachial spinal cords, controlling limb innervation of distal nerves (Catela, Chen, Weng, Wen, and Kratsios, 2022). Similar regulatory pathways involving Hox/Mes in Ret expression have been suggested at lumbar levels, with misexpression of Hoxc10 or Hoxd10 in chicken embryos resulting in Ret expression arising in the neural tube (Catela et al., 2022). Our bulk RNA-seq analysis identified Hoxc11 as one of the DEGs, indicating a potential compromise of proper lumbar identity in the absence of Isl2. However, the expression of Ret and Gfra1 was not changed in Isl2 KO mice when assessed by bulk RNA-seq analysis and in situ hybridization, and overall hindlimb motor projections remained intact. Although further investigation is necessary to fully elucidate the hierarchy among Hox, GDNF and Isl2, it is evident that each factor plays a crucial role, to varying degrees, in the proper expression of Etv4 within proximal hindlimb motor pools.”

4) Improve quantifications: The way MN numbers are given throughout the manuscript is confusing. The data should be recalculated and presented as average MN numbers/mouse (n=1 mouse and not 1 section). In Figure 2, quantification would be really good to have at both stages. The statistical difference in MMC and PGC numbers (Figure 2f) observed between WT and KO mice should be briefly commented on, in light of Thaler's results and in the context of the ectopic motor neurons.

We re-analyzed the motor neuron count, presenting it as an average number of motor neurons per section derived from individual embryos of each genotype across Figure 2 to Figure 7. We also provide a succinct comparison of our quantification outcomes with those previously reported by Thaler et al.

Page 6: “Our findings align with previous results, including a reduced number of visceral motor neurons, reduced nNOS expression, and abnormal dorsomedial migration of motor neurons, reported by Thaler et al. Our study additionally reveals the appearance of ectopic motor neurons at the limb levels, primarily consisting of MMC and LMCl cells.

We have included cell count of ectopic motor neurons in embryos in Figure 2—figure supplement 1c. The quantification of adult motor neurons has been relocated to Figure 7 to improve the continuity. Since the columnar identity is less discernible in adult motor neurons, our quantification focused solely on determining the total motor neuron count within each segment (Figure 7b). Furthermore, our investigation of the medio-lateral distribution of ectopic motor neurons has revealed the presence of such neurons in Isl2 null mice (Figure 7c).”

Page 11: “Overall, the total number of ChAT+ motor neurons remained unchanged, and mislocalized motor neurons were observed in the P21 Isl2-null lumbar spinal cord (Figure 7a-c).”

5) Figure 3: The markers used in Figure 3a do not allow to clearly distinguish MMC, LMCm and LMCl neurons. A different combination should be used such as Lhx3 (MMC) Isl1 (LMCm) and Hb9 (LMCl). In addition, spatial analysis of columnar and divisional segregation should be provided to better support the statement.

To better distinguish motor columns, we used additional markers Hb9, Lhx3, Lhx1 for our analysis in Figure 3 as the Reviewer suggested. sMN (Hb9), MMC (Hb9, Lhx3), HMC (Hb9, Scip), LMCm (Hb9low, Isl1, Foxp1), LMCl (Hb9high, Lhx1, Foxp1) were carefully defined using these markers. Further, FCU (Scip), CM (Etv4, Isl1), LD (Etv4, Lhx1) motor pools were defined as well. Contour, density and spatial plots of motor columns were provided in Figure 3-supplement 2 and 3.

6) Further experimentation and/or discussion is needed to clarify which exact motor pools are affected in Isl2 mutant mice. See comments by reviewers 1 and 3.

We sincerely appreciate the valuable feedback offered by the Reviewer. Following the suggestion of our Reviewer, we expanded our study to encompass the developmental examination of other Etv4-expressing motor pools, namely Rf and Tfl, in addition to Gl. Our comprehensive analysis uncovered noteworthy developmental impacts on the Tfl motor pools, while the Rf motor pools remained unaffected. Considering the divergence of Tfl axons from Gl projections, it is plausible that these motor pools share analogous developmental genetic programs. A previous study reported that Etv4 expression is transient in Rf motor pools, lasting until E13.5 (De Marco Garcia and Jessell, 2008). Thus, the postnatal development of Rf motor pools could be independent of Etv4.These results support the notion that Isl2 exerts both a universal and distinct influence on Etv4+ motor pools, albeit to varying extents. We have supplemented our findings with illustrative images and quantification data (Figure 7d, Figure 8h-j and Figure 8—figure supplement 3), providing further elaboration within the Results and Discussion sections of our manuscript.

Page 11: “Similarly, Tfl motor pools, whose axons diverged from one of the major branches of Gl ones, also displayed shorter and lesser dendritic arbors in Isl2 KO mice (Gould et al., 2008). The position and dendritic arborization of Rf motor pools remained unchanged in Isl2 KO mice (Figure 7d). Considering that Etv4 expression has been previously reported to be transient until E13.5 in Rf motor pools, it may not play a significant role in the dendritic arborization of Rf motor pools (De Marco Garcia and Jessell, 2008).”

Page 13: “In Tfl muscle, abnormal AChRs were formed along axon branches at P0, and a drastic loss in AChR clusters and polyinnervation were observed at P28 in Isl2 KO mice (Figure 8h, i and j). However, Rf muscles, whose motor pools transiently express Etv4, showed normal arrangement of NMJs (Figure 8—figure supplement 3) (De Marco Garcia and Jessell, 2008).”

Page 15: “The removal of GDNF, its receptor Ret or Gfra1, or Etv4 has been mostly characterized in CM and LD motor pools, the Etv4-expressing motor pools at the forelimb level, but their roles in the lumbar motor pools have not been addressed (Arber et al., 2000; Haase et al., 2002; Livet et al., 2002; Vrieseling and Arber, 2006). In our study, we observed a selective loss of Etv4 expression only in the lumbar spinal cord of Isl2 KO mice, specifically in Gl and Tfl motor pools. These motor pools displayed a phenotype almost identical to CM and LD motor pools of Etv4-null mice, including mispositioned cell bodies, impaired dendrite and axon arborization, defective NMJs, and reduced sensorimotor connectivity. In addition, acute knockdown of Isl2 in the chicken neural tube and in motor neuron-specific Isl2 cKO mice was sufficient to abolish Etv4 transcripts, suggesting that Isl2 positively regulates Etv4 expression in a cell-autonomous manner. Thus, GDNF likely plays broader roles in guiding most LMC projections and also promoting differentiation of Etv4-expressing motor pools at both limb levels. In contrast, Isl2 plays a precise role in fine-tuning the position of motor neurons, including MMC and LMCl, and also functions in selected lumbar motor pools at the hindlimb, particularly in establishing sensorimotor connectivity during the late period of development.”

Reviewer #1 (Recommendations for the authors):

Figure 2 is confusing. The presentation and discussion of the data in Results (lines 149-162) need to be improved. The authors should also present their data more clearly in the context of Thaler et al. (2004), as that study found increased numbers of somatic motor neurons and decreased numbers of visceral motor neurons in the thoracic spinal cord. By doing so, the novelty of the current study will be clearer. Perhaps, switching the order of Figures 2 and 3 would address this issue and improve the flow.

Thank you for the valuable feedback from the Reviewer. We re-analyzed number of motor neurons as an average number of motor neurons per section derived from one embryo of each genotype in Figure 2 to Figure 7. Thaler et al. presented Hb9+ motor neuron counts across spinal cord segment in a histogram, showing a slightly greater count in Isl2 null mice. However, they did not conduct statistical analysis on this. In contrast, our study separately quantified MMC and HMC, averaging their numbers along the thoracic segment. According to these criteria, the number of MMC and HMC in Isl2 KO mice did not significantly differ. We also briefly discuss about our quantification results compared to the previous one done by Thaler et al. as follows.

Page 6: “Our findings align with previous results, including a reduced number of visceral motor neurons, reduced nNOS expression, and abnormal dorsomedial migration of motor neurons, reported by Thaler et al. Our study also reveals the appearance of ectopic motor neurons at the limb levels, primarily consisting of MMC and LMCl cells.”

As Figure 3 and 4 collectively present the motor column distribution, we have chosen to retain the initial sequence of Figure 2 and 3. In this revised arrangement, the outcomes pertaining to adult animals (Figure 2e) have been shifted to Figure 7a-c to ensure a smoother transition.

In the transcriptomic analysis, the authors do a remarkable job to identify motor neuron-expressed genes and define subclusters of LMC motor pools. Although the RNA-Seq is done in bulk and without motor neuron specificity, the authors leverage published RNA-Seq data to ascribe MN specificity to their dataset. Hence, they identified several putative targets, which they validate with ISH (Figure 6H). However, they are strongly encouraged to also validate with ISH putative markers of the subcluster LMCl.2 in Isl2 global or conditional mice. This experiment would significantly strengthen the authors' conclusions (lines 356-366) about which motor pools are affected by Isl2 loss, especially because Pea3 is a marker for multiple motor pools.

We have presented ISH results of six DEGs that showed downregulation in Isl2 null mice. Remarkably, five of these genes (C1al3, Etv4, Kcnab1, Prph and Anxa2) were among the Top 30 enriched genes in LMCl.2 (Figure 5h and Supplementary Table S3). In response to the Reviewer’s recommendation, we have additionally included additional ISH results for LMCl.2 markers-Epha3, Fgf10 and Sema5a. Our observations indicated that the transcripts of these markers were exclusively present in Gl motor pools, with their expression levels being diminished in Isl2 null mice (Figure 6i).

Page 10: For instance, Epha3, Fgf10, and Sema5a transcripts were enriched in Gl motor pools and their expression was reduced in Isl2 KO, as validated by in situ hybridization analysis (Figure 6i, Supplementary Table S3). In conclusion, our integrated approach, combining bulk and scRNA-seq analyses datasets, has proven valuable in identifying new potential motor pool markers and uncovering novel transcription target genes.”

The study will be strengthened by conducting hindlimb locomotion analysis in Olig2Cre;Isl2 fl/fl mice.

Regrettably, Isl2 conditional mice experience mortality within a day after birth, precluding the possibility of conducting behavioral studies. A comprehensive presentation of the outcomes and their interpretation concerning the observed disparity in the mortality of these two mouse lines has been detailed. These details can be referenced in in Table S1, Supplementary Videos 1-3, as well as within the Results and Discussion section (pages 10, 16).

The study will be strengthened by conducting hindlimb locomotion analysis in Olig2Cre;Isl2 fl/fl mice.

The text can be significantly improved by eliminating typos and improving grammar. The authors should go carefully through the text, sentence by sentence, and make the necessary changes.

We thank to the Reviewers for bringing our attention to the grammatical errors. In order to improve our manuscript, it has undergone a meticulous editorial process by a professional editor.

In the first section of Results, the authors can cite Thaler et al. (2004), as their results are consistent with that study.

We have cited the reference as the Reviewer suggests in p 6, line 132, 135.

Reviewer #2 (Recommendations for the authors):

In my opinion, the writing is at times not very clear and makes it difficult to appreciate the findings and their relevance in the context of the literature. The experiments are well executed and appropriately analyzed, there are a few things that can be improved, more to it in the following detailed comments:

1) I find the introduction difficult to read, it is more like a list of facts than a coherent narrative to properly frame the work. For example, the bit about neuromesodermal progenitors (lines 66-71), while interesting does not fit with the rest of the story. Also, at times the balance between subjects seems a bit off, more emphasis should be given to what is known about Isl2 and Pea3 mutant mice, which is very relevant to the paper.

In order to enhance the clarity of our findings and accentuate key aspects of our study, we have made the decision to eliminate certain less pertinent sections. Simultaneously, we have enriched the manuscript by incorporating a more comprehensive overview of prior investigations involving mutant mice of Isl2 and Pea3. The details of these enhancements are elaborated below.

Page 3: “More than 50 limb muscles in tetrapods are innervated by distinct limb motor pools, suggesting the involvement of diverse molecular mechanisms in shaping individual limb motor pools during spinal cord development (Sullivan, 1962). Specific LMC motor pools express ETS factors Etv4 and Etv1, playing a role in acquiring motor pool identities. In Etv4 mutant mice, the position and terminal arborization of specific motor pools were perturbed, resulting in disorganized motor pools and impaired motor control (Livet et al., 2002; Vrieseling and Arber, 2006). GDNF is suggested to act as a peripheral signal derived from limb tissues, guiding major axon bundles toward the hindlimb (Gould et al., 2008; Kramer et al., 2006). Furthermore, GDNF signaling appears to induce Etv4 expression in certain motor pools, as deletion of gdnf, its receptor Ret, or Gfra1 results in extinguished Etv4 expression and affected terminal arborization of motor axons in target muscles (Haase et al., 2002; Livet et al., 2002). However, it remains unclear whether the development of Etv4-expressing motor pools solely depends on GDNF signaling.”

2) Line 150: it is not specified which Isl2 KO mice is used. At the very least there should be a reference (Thaler et al., 2004). However, here it is very important to comment that these are the same mice that have been used in a previous study characterizing the role of Isl2 in spinal motor neuron development. In addition, the mice were originally reported to "die within 24 hr of birth" (Thaler et al., 2004), but the authors found that "approximately 10% survived" (lines 315-317). This important information should be introduced at the beginning when the mouse line is first introduced. It will be immediately relevant as in line 160 adult isl2 KO mice are used.

In order to provide contextual clarity, we have integrated the original reference and added a statement clarifying the utilization of the mutant mouse line established and analyzed by Thaler et al. in our study (Page 6). Comprehensive insights and explanations pertaining to the disparity in mortality between the two mouse lines are thoroughly presented and accessible in Table S1, Supplementary Videos 1-3, and the Results and Discussion section (pages 10, 16).

Page 6: “To explore unknown phenotypes, we conducted further investigations using the mutant mouse line previously generated and analyzed by Thaler et al. (Thaler et al., 2004).”

3) Line 159: "Mislocalized motor neurons remained in the adult Isl2-null lumbar spinal cord with a similar number of motor neurons" Mislocalised motor neurons are not very evident from pictures provided in figure 2h higher magnifications would be helpful. Regarding the "similar number", I could not find quantification of mislocalised motor neurons either in embryos or adults, thus a quantitative statement is not justified. Indeed, quantification would be really good to have at both stages.

We refined the original figures by adjusting their cropping to enhance the visibility of ectopic motor neurons. Additionally, for improved continuity, we re-located figure 2h to figure 7a. To offer a precise quantification of ectopic motor neurons, we have incorporated the cell count results within the text and figures 7b, c.

Page 11: “Overall, the total number of ChAT+ motor neurons remained unchanged, and mislocalized motor neurons were observed in the P21 Isl2-null lumbar spinal cord (Figure 7a-c).”

4) The statistical difference in MMC and PGC numbers (Figure 2f) observed between WT and KO mice should be briefly commented on, in light of Thaler's results and in the context of the ectopic motor neurons.

We have added some sentences to interpret our quantification results of motor neurons together with Thaler’s results as below.

Page 6: “Our findings align with previous results, including a reduced number of visceral motor neurons, reduced nNOS expression, and abnormal dorsomedial migration of motor neurons, reported by Thaler et al. Our study also reveals the appearance of ectopic motor neurons at the limb levels, primarily consisting of MMC and LMCl cells.”

5) Line 169: "We similarly found that MMC neurons were scattered dorsolaterally and located ectopically, either within the LMC columns or immediately dorsal to the motor columns in Isl2-deficient (Figure 3a)" The markers used in Figure 3a do not allow to clearly distinguish MMC, LMCm and LMCl neurons. A different combination should be used such as Lhx3 (MMC) Isl1 (LMCm) and Hb9 (LMCl). In addition, spatial analysis of columnar and divisional segregation should be provided to better support the statement.

To better distinguish motor columns, we used additional markers Hb9, Lhx3, Lhx1 for our analysis in Figure 3 as the Reviewer suggested. sMN (Hb9), MMC (Hb9, Lhx3), HMC (Hb9, Scip), LMCm (Hb9low, Isl1, Foxp1), LMCl (Hb9high, Lhx1, Foxp1) were carefully defined using these markers. Further, FCU (Scip), CM (Etv4, Isl1), LD (Etv4, Lhx1) motor pools were defined as well. Contour, density and spatial plots of motor columns were provided in Figure 3-supplement 2 and 3.

6) Figure 4b: I believe there must be a mistake as it is clear from the pictures in 4a that there are very few Pea3+ neurons in Isl2 KO mice, however in the plots in 4b (last row the three plots on the right) there are huge red contours for Pea3 neurons more or less similar to control mice. Indeed, the difference is clearly evident in the dot plots in Figure 4 supp 1b.

Initially, we employed an R package to generate contour plots, but this approach neglected the cell density aspect. This resulted in misleading interpretations, suggesting a higher presence of Etv4+cells in Isl2 KO mice. To rectify this, we have optimized the code to incorporate cell density considerations and subsequently re-created all contour plots in Figure 1, Figure 3—figure supplement 2 and Figure 4—figure supplement 1. To enhance visualization and facilitate accurate comparisons, we have further improved the plots by overlaying the position of cells with Etv4+ cell contours (Figure 4—figure supplement 1).

7) It is great that the authors generated an Isl2 conditional allele, however, its phenotype in combination with olig2::cre is not described other than in the experiments in Figure 5 aimed at studying the downregulation of Pae3. Does the Isl2 cKO mouse recapitulate the overall defects observed in Isl2 KO in terms of motor neuron generation, spatial organisation, connectivity, and behaviour? Comparison of the two can be very informative especially regarding the behavioural and sensorimotor connectivity phenotypes as many other cells (i.e., all somatosensory neurons) express Isl2 during development.

Initially, our experiments primarily involved the use of global Isl2 KO mice, and more recently, we generated Isl2 conditional mice to address the issue of cell autonomy. Through meticulous observation, we identified that about 10% of global Isl2 KO survived, allowing us to effectively examine their postnatal phenotypes. Regrettably, the Isl2 conditional KO (cKO) mice showed a different survival rate, as they all succumbed within a day after birth. Thus, our ability to perform experiments utilizing postnatal animals was limited.

Detailed results and the interpretation surrounding the observed mortality discrepancy in these two mouse lines have been comprehensively presented in Table S1, Supplementary Videos 1-3, and elaborated within the Results and Discussion section.

Nevertheless, we successfully investigated the progress of NMJ development in Isl2 cKO at E18.5, just prior to birth, and identified congruent developmental anomalies comparable to Isl2 null mice. Consequently, the phenotypes we observed during the embryonic stage were identical in both mouse lines, i.e., dispersed motor neurons, Etv4 expression absence in selective lumbar motor pools, disrupted NMJ formation within those pools. The new obtained data, along with pertinent explanations, have been integrated into Figure 8—figure supplement 2 and incorporated within the Results and Discussion section, as outlined below.

Page 12: “Due to the postnatal lethality of Isl2 cKO, we analyzed the muscles of these mice at E18.5, prior to birth, when NMJ formation is in progress (Figure 8—figure supplement 2). GFP-labeled axons arrived in Isl2-deficient Gl muscles; however, their terminal arbors were defasciculated and misrouted. Fluorescence intensity of α-bungarotoxin (BTX) at the NMJs was greatly reduced in Isl2 cKO. Furthermore, AChR clustering appeared to be incomplete, showing slender and elongated NMJs, or some NMJs were abnormally concentrated in proximity to the primary axon shaft in Isl2 cKO (Figure 8—figure supplement 2b).”

8) Do you observe arborization, sensorimotor connectivity, and NMJ defects shown for Gl motor neurons also in Rf and Tfl motor pools? It seems like it would be the case but it would be a great addition in order to generalise the role of Isl2-Pea3 transcriptional axis on all pools expressing it.

We sincerely appreciate the valuable feedback offered by the Reviewer. Following the suggestion of our Reviewer, we expanded our study to encompass the developmental examination of other Etv4-expressing motor pools, namely Rf and Tfl, in addition to Gl. Our comprehensive analysis uncovered noteworthy developmental impacts on the Tfl motor pools, while the Rf motor pools remained unaffected. Considering the divergence of Tfl axons from Gl projections, it is plausible that these motor pools share analogous developmental genetic programs. A previous study reported that Etv4 expression is transient in Rf motor pools, lasting until E13.5 (De Marco Garcia and Jessell, 2008). Thus, the postnatal development of Rf motor pools could be independent of Etv4. These results support the notion that Isl2 exerts both a universal and distinct influence on Etv4+ motor pools, albeit to varying extents. We have supplemented our findings with illustrative images and quantification data (Figure 7d, Figure 8h-j and Figure 8—figure supplement 3), providing further elaboration within the Results and Discussion sections of our manuscript.

Page 11: “Similarly, Tfl motor pools, whose axons diverged from one of the major branches of Gl ones, also displayed shorter and lesser dendritic arbors in Isl2 KO mice (Gould et al., 2008). The position and dendritic arborization of Rf motor pools remained unchanged in Isl2 KO mice (Figure 7d). Considering that Etv4 expression has been previously reported to be transient until E13.5 in Rf motor pools, it may not play a significant role in the dendritic arborization of Rf motor pools (De Marco Garcia and Jessell, 2008).”

Page 13: “In Tfl muscle, abnormal AChRs were formed along axon branches at P0, and a drastic loss in AChR clusters and polyinnervation were observed at P28 in Isl2 KO mice (Figure 8h, i and j). However, Rf muscles, whose motor pools transiently express Etv4, showed normal arrangement of NMJs (Figure 8—figure supplement 3) (De Marco Garcia and Jessell, 2008).”

Reviewer #3 (Recommendations for the authors):

Specific concerns:

1) The behavioral phenotypes seen in the videos appear to be very dramatic, especially in the hindlimbs. However, the authors show that most hindlimb muscles are properly innervated, with the exception of the Gl. For example, the Rf seems to be properly innervated but its weight is dramatically reduced in Figure 9. What is the cause for this? Do the authors believe that the behavior correlates to defects in Gl development and connectivity or are there non-MN contributions to these phenotypes from sensory neurons and interneurons?

Our observations revealed a significant disruption in the development of both Gl and Tfl motor pools, manifesting in altered neuron branching patterns and impaired NMJ formation. Given the substantial role of Gl and Tfl muscles as hip extensors and supporters of hip movements, their compromised function could adequately explain the observed rigidity in hip movement within Isl2 null mice. Although the development of Rf motor pools was relatively less affected, we postulate that the reduced muscle weight in Rf could be attributed to a broader hindlimb movement reduction.

To delineate whether the observed defects in global Isl2 KO mice were solely attributed to motor neurons, we conducted further investigations by characterizing motor neurons in Isl2 cKO, where Isl2 was specifically absent within motor neurons. Intriguingly, the phenotypes we identified in these two mouse lines were strikingly similar, encompassing dispersed motor neurons, the absence of Etv4 expression in selective lumbar motor pools, and disrupted NMJ formation. Further, our investigations extended to the chicken neural tube, where acute knockdown of Isl2 likewise abolished Etv4 expression. Detailed information, including the new data, has been incorporated in Figure 8—figure supplement 2 and incorporated them into the Results and Discussion section as below.

Page 12: “Due to the postnatal lethality of Isl2 cKO, we analyzed the muscles of these mice at E18.5, prior to birth, when NMJ formation is in progress (Figure 8—figure supplement 2). GFP-labeled axons arrived in Isl2-deficient Gl muscles; however, their terminal arbors were defasciculated and misrouted. Fluorescence intensity of α-bungarotoxin (BTX) at the NMJs was greatly reduced in Isl2 cKO. Furthermore, AChR clustering appeared to be incomplete, showing slender and elongated NMJs, or some NMJs were abnormally concentrated in proximity to the primary axon shaft in Isl2 cKO (Figure 8—figure supplement 2b).”

2) The connection between Isl2/gdnf signaling and Pea3 expression is not experimentally supported. The authors do show a specific downregulation of Pea3 at lumbar levels of the spinal cord but whether this is mediated through gdnf is not established. Is Ret or any other component of this signaling pathway downregulated in the RNA-seq data? The fact that lumbar explants do not respond to gdnf may suggest this but the connection is not explicitly shown.

We sincerely appreciate the valuable feedback provided by the Reviewer. We have integrated a pertinent description within the Discussion section (Pages 15-16) to elucidate the connection between Isl2/gdnf signaling and Pea3 expression. We also have included additional data illustrating Ret and Gfra expression in Isl2 null mice (Figure 8—figure supplement 1), along with relevant interpretations presented in the Discussion section (Page 16). We readily acknowledge the Reviewer’s concerns regarding the limitations of the explant experiment, which encompassed a mixture of Pea3-expressing and non-expressing LMC motor axons. Consequently, we have made the necessary adjustment by excluding the in vitro data and, instead, dedicated our efforts to an in-depth discussion of the disparities observed between mice lacking gdnf signaling and those with Isl2 mutations in our study.

Page 15: “Additional compelling evidence that Isl2 mainly functions in Etv4+ motor pools comes from the striking similarities and differences observed in the mutant phenotypes of Isl2 and the GDNF-Etv4 pathway, requiring a detailed comparison. GDNF has been recognized as a critical factor for the survival and innervation of motor neurons during early motor neuron development (Henderson et al., 1994). Deletion of GDNF or its receptor Ret or Gfra1 leads to impaired innervation of the peroneal nerve, resulting in no or disrupted axon innervation in major hindlimb muscles, such as Edl, Atib, psoas major, Gmax, and Tfl, to varying degrees (Bonanomi et al., 2012; Gould et al., 2008; Kramer et al., 2006). In contrast, Isl2 mutant mice showed normal axon projection of the peroneal nerve and relatively normal axon innervation in most hindlimb muscles, unlike motor axons of mice deficient in GDNF or its receptors. Notably, discrepancies between the roles of GDNF and Isl2 also emerged in types of motor pools affected. The removal of GDNF, its receptor Ret or Gfra1, or Etv4 has been mostly characterized in CM and LD motor pools, the Etv4-expressing motor pools at the forelimb level, but their roles in the lumbar motor pools have not been addressed (Arber et al., 2000; Haase et al., 2002; Livet et al., 2002; Vrieseling and Arber, 2006). In our study, we observed a selective loss of Etv4 expression only in the lumbar spinal cord of Isl2 KO mice, specifically in Gl and Tfl motor pools. These motor pools displayed a phenotype almost identical to CM and LD motor pools of Etv4-null mice, including mispositioned cell bodies, impaired dendrite and axon arborization, defective NMJs, and reduced sensorimotor connectivity. In addition, acute knockdown of Isl2 in the chicken neural tube and in motor neuron-specific Isl2 cKO mice was sufficient to abolish Etv4 transcripts, suggesting that Isl2 positively regulates Etv4 expression in a cell-autonomous manner. Thus, GDNF likely plays broader roles in guiding most LMC projections and also promoting differentiation of Etv4-expressing motor pools at both limb levels. In contrast, Isl2 plays a precise role in fine-tuning the position of motor neurons, including MMC and LMCl, and also functions in selected lumbar motor pools at the hindlimb, particularly in establishing sensorimotor connectivity during the late period of development.

Nevertheless, Etv4 expression in the CM motor pool at the brachial level was intact in both Isl2 KO mice and cKO mice, implying that another complex genetic program works in conjunction with Isl2 to contribute to segment-specific regulation. For instance, multiple Hox genes control GDNF signaling, with Hoxc6 activating Ret and Gfra1 expression in rostral brachial spinal cords, regulating terminal arborization of CM and LD motor pools, while Hoxc8 activates Ret and Gfra3 expression in caudal brachial spinal cords, controlling limb innervation of distal nerves (Catela et al., 2022). Similar regulatory pathways involving Hox/Mes in Ret expression have been suggested at lumbar levels, with misexpression of Hoxc10 or Hoxd10 in chicken embryos resulting in Ret expression arising in the neural tube (Catela et al., 2022). Our bulk RNA-seq analysis identified Hoxc11 as one of the DEGs, indicating a potential compromise of proper lumbar identity in the absence of Isl2. However, the expression of Ret and Gfra1 was not changed in Isl2 KO mice when assessed by bulk RNA-seq analysis and in situ hybridization, and overall hindlimb motor projections remained intact. Although further investigation is necessary to fully elucidate the hierarchy among Hox, GDNF and Isl2, it is evident that each factor plays a crucial role, to varying degrees, in the proper expression of Etv4 within proximal hindlimb motor pools.”

3) The loss of synapses in Fig7b is interpreted as Isl2 controlling Gl connectivity. Since this is not a conditional allele, the synaptic loss could be due to defects in sensory neurons. The authors also show an ectopic soma (although this is not really well defined). Is there a difference between ectopic and Gl MNs that settle in the correct position? Are the authors suggesting that cell body position influences connectivity in a way that ectopic MNs lose more connections? This is not quantified or discussed.

We express our gratitude for the valuable comments provided by the Reviewer. Regrettably, the postnatal lethality exhibited by Isl2 cKO mice precluded our ability to explore Gl connectivity in this context. However, the observed NMJ defects in Isl2 cKO animals substantiate the pivotal role of Isl2 within motor neurons. In alignment with the Reviewer’s guidance, we undertook the categorization of motor neurons into two distinct zones: zone 1 encompassing motor neurons positioned correctly, and zone 2 encompassing ectopic somata. Our findings unveiled a trend wherein ectopic somata tended to exhibit a lower number of sensory boutons per soma, suggesting a clear influence of cell body positioning on connectivity. To provide a comprehensive representation, we have incorporated pertinent data within Figure 7e-k and incorporated them into the Results section as follows.

Page 11: “Next, we analyzed the synaptic density of vGluT1+ boutons on motor neuronal somata. Overall, the density of vGluT1 immunoreactivity in zone 1 was higher than in zone 2, which is expected since more motor neurons reside in zone 1 where more sensorimotor synapses are present. Compared to the control mice, the density of vGluT1 immunoreactivity in Isl2 KO mice did not change, indicating that overall sensorimotor connection is not affected in them (Figure 7h). However, when we analyzed the number of vGluT1+ boutons in contact with soma of Rh-Dex+ Gl motor neurons, we observed a great reduction in the number of boutons in Isl2 KO mice (Figure 7i-k). Furthermore, the number of boutons in zone 2, but not in zone 1, was significantly reduced in Isl2 KO mice (Figure 7k). Thus, ectopically located Gl motor neurons are more likely to receive fewer proprioceptive sensory inputs compared to the motor neurons in a normal position in the absence of Isl2. Collectively, our findings suggest that Isl2 KO mice displayed reduced and misoriented dendrites of Gl motor pools, leading to defective sensorimotor connections.”

4) The way MN numbers are given throughout the manuscript is confusing. The authors quantify MN/section. However, especially for motor pool counts, this results in a lot of variability since at the beginning and end of each motor pool you have 1-2 MNs, while higher numbers are in the middle. The data should be recalculated and presented as average MN numbers/mouse (n=1 mouse and not 1 section). This would reduce variability within the datasets.

We re-analyzed number of motor neurons as an average number of motor neurons per section derived from one embryo of each genotype in Figure 2 to Figure 7.

5). Figure 1: The authors discuss the different spatiotemporal expressions of Isl1 and Isl2 but this is not shown (maybe inferred from RNAseq). Only one time point is shown in the figure (e12.5). It would be nice to show the temporal expression of both TFs at different embryonic time points and also in combination with motor pool markers. For example, the overlap of Isl2 with Pea3 at brachial and lumbar levels is not shown. Showing which motor pools express high and sustained levels of Isl2 in this figure would be a nice setup for the rest of the story.

To refine the temporal depiction of Isl2 and Pea3 expression dynamics, we have enriched Figure 1d and e with contours or additional images. These results illuminate the spatiotemporal distribution of Isl1, Isl2 and Etv4 at E10.5 and/or E12.5. Furthermore, we have provided further details in Result section as below. In brief, our observations delineate the initiation of Isl2 expression in early postmitotic motor neurons, slightly lagging behind Isl1 across all motor neurons. Meanwhile, Etv4 expression remains confined to specific LMC motor pools.

Page 5: “At E10.5, during motor neuron generation, Olig2+ pMNs were initially found in the middle part of the ventral spinal cord, co-expressing Lhx3, a marker for pMNs and MMC (Lee, Lee, Ruiz, and Pfaff, 2005; Thaler, Lee, Jurata, Gill, and Pfaff, 2002) (Figure 1d). As these cells migrated, they transiently expressed markers for postmitotic motor neurons Isl1 and Hb9. Isl2 expression emerged slightly later, found in more lateral regions across all levels analyzed. To analyze the distribution of Isl2 during the motor neuron maturation, we compared the relative position of three populations: Olig2+ pMNs, Olig2+Isl1+ immature motor neurons, and Isl2-expressing motor neurons. Contour and density assessments revealed that Isl2 expression began slightly after Isl1 expression in immature motor neurons and persisted in postmitotic motor neurons along with Isl1 (Figure 1d). At E12.5, when motor columns became distinct, we identified them based on the expression of motor neuronal markers, including Foxp1 (marker for LMC), Isl1 (marker for LMCm and PGC), Lhx3 (marker for MMC), nNOS (marker for PGC), and Etv4 (marker for some LMC motor pools). Motor pools expressing Etv4 were located in the LMC but not at the thoracic level (Figure 1e). Isl2 exhibited broad expression across all motor columns, unlike other LIM-HD transcription factors that showed specific expression patterns. Overall, while Isl2 is present in differentiated motor neurons, it is relatively enriched in MMC and LMCl populations, where its expression is more prominent compared to Isl1.”

6). Figure 2-4: MN numbers should be recalculated and presented so that n=1 mouse. It is likely that this might change some of the data. For example, in Figure 3a, there is a visible reduction of Scip+ MNs. When the source data for Figure 3c was reanalyzed, there seems to be a 30% reduction of FCU neurons with a P=0.03 (not significant but close, perhaps would be significant if n>3). Is Isl2 highly expressed in Scip+ MNs at brachial levels (see point for Figure 1). Also the MN positional changes should be quantified in these figures.

We re-analyzed number of motor neurons as an average number of motor neurons per section derived from one embryo of each genotype in Figure 2 to Figure 7. We replaced the image of Scip neurons, which provides a more representative depiction of their average distribution in C6 and C8 segments. We also reanalyzed FCU counts as the average number of MNs per embryo, revealing no significant difference. In addition, we provided density plots and boxplots showing distribution of neurons along V-D and M-L axes in Figure 3 and 4.

7). Figure 6: After comparing their RNAseq data to scRNAseq the authors mention that only 50% seem to be MN-specific. Is this technical (not deep enough sequencing in the scRNA-seq datasets) or are the rest of the genes in interneurons? This relates to the comment about other affected populations in the null mouse. In the in situ data, the ventrolateral pools are indicated, however, the changes appear to be broader, affecting the entire lumbar spinal cord. The nNos data should be shown separately since that is in the thoracic spinal cord.

In our investigation, we conducted bulk RNAseq analysis on tissue samples from the ventral part of embryonic spinal cord tissues. Thus, the analyzed transcripts could potentially originate from either motor neurons or other cell types. To pinpoint genes within motor neurons, we referred to the list of genes established to exist in motor neurons. Employing a threshold filter, we considered genes with a minimum raw read count greater than 1 per gene from the scRNA-seq dataset obtained from embryonic motor neurons. Such threshold filters are commonly used in transcriptomic analyses, often with varying read count threshold ranging from 1 to around 10, contingent upon specific research criteria (Corchete et al., 2020; Evrony, Lee, Park, and Walsh, 2016; Namboori et al., 2021). To capture potential DEGs affiliated with motor neurons without losing valuable candidates, we adopted a minimum raw read count > 1 per gene. However, given that our data emanated from the entire ventral spinal cord and did not further refine genes exclusively expressed in motor neurons, some candidate genes showed broader expression. Regardless of the origin of the transcript, whether motor neuron or broadly expressed, we verified that the levels of transcripts were diminished in specific motor pools of interest in Isl2 mutant mice, in alignment with our DEG analysis. In our revised manuscript, we have spotlighted ISH results from six DEGs and an additional three enriched genes identified from scRNAseq analyses. These genes collectively exhibit relatively specific expression in Gl motor pools, with a noticeable distinction of its transcript in Isl2 null mice (Figure 6h, i). Lastly, we have removed nNos data since it is already reported earlier in Figure 2.

Page 10: For instance, Epha3, Fgf10 and Sema5a transcripts were enriched in Gl motor pools and their expression was reduced in Isl2 KO, validated by in situ hybridization analysis (Figure 6i, Supplementary Table S3). Collectively, our integrated approach, combining bulk and scRNAseq analyses datasets, proves valuable in identifying new potential motor pool markers and uncovering novel transcription target genes.”

8). Figure 7: In the text, it is mentioned that the Pea3 pools are Gl, Rf, and Tfl and that a retrograde tracer was injected into individual muscles. However, only the Gl data are shown. What about the other pools? Do they have normal dendrites? It seems that Pea3 is not the whole story as Rf has normal NMJ innervation. What about sensory innervation? Is that synapse loss specific to Gl MNs? What about Rf and Tfl? What about non-Pea3+ pools?

We sincerely appreciate the valuable feedback offered by the Reviewer. Following the suggestion of our Reviewer, we expanded our study to encompass the developmental examination of other Etv4-expressing motor pools, namely Rf and Tfl, in addition to Gl. Our comprehensive analysis uncovered noteworthy developmental impacts on the Tfl motor pools, while the Rf motor pools remained unaffected. Considering the divergence of Tfl axons from Gl projections, it is plausible that these motor pools share analogous developmental genetic programs. A previous study reported that Etv4 expression is transient in Rf motor pools, lasting until E13.5 (De Marco Garcia and Jessell, 2008). Thus, the postnatal development of Rf motor pools could be independent of Etv4. These results support the notion that Isl2 exerts both a universal and distinct influence on Etv4+ motor pools, albeit to varying extents. We have supplemented our findings with illustrative images and quantification data (Figure 7d, Figure 8h-j and Figure 8—figure supplement 3), providing further elaboration within the Results and Discussion sections of our manuscript.

Page 11: “Similarly, Tfl motor pools, whose axons diverged from one of the major branches of Gl ones, also displayed shorter and lesser dendritic arbors in Isl2 KO mice (Gould et al., 2008). The position and dendritic arborization of Rf motor pools remained unchanged in Isl2 KO mice (Figure 7d). Considering that Etv4 expression has been previously reported to be transient until E13.5 in Rf motor pools, it may not play a significant role in the dendritic arborization of Rf motor pools (De Marco Garcia and Jessell, 2008).”

Page 13: “In Tfl muscle, abnormal AChRs were formed along axon branches at P0, and a drastic loss in AChR clusters and polyinnervation were observed at P28 in Isl2 KO mice (Figure 8h, i and j). However, Rf muscles, whose motor pools transiently express Etv4, showed normal arrangement of NMJs (Figure 8—figure supplement 3) (De Marco Garcia and Jessell, 2008).”

To discern whether the defects evident in the global Isl2 KO mice stem exclusively from motor neurons, we undertook a detailed characterization of motor neurons in Isl2 cKO, where Isl2 was specifically lacking within motor neurons. Notably, the observed phenotypes in both mouse lines was remarkably similar, i.e., dispersed motor neurons, the absence of Etv4 expression in selective lumbar motor pools, disrupted NMJ formation. This congruence in outcomes was further underscored by our observations in the chicken neural tube, where acute knockdown of Isl2 led to the distinction of Etv4 expression.

Additionally, we investigated the spindle innervation of sensory Ia to Gl muscles in Isl2 null mice. The results indicated an unaffected sensory nerve innervation, suggesting that Isl2 deficiency did not influence sensory nerve connections. It is important to note that, given the preliminary nature of this finding and its relevance ot ongoing studies, we have opted to present these results exclusively for the Reviewers at this stage (see Author response image 1b). Also we examined dendrite patterning in Ta motor pools as an example of non-Pea3 motor pools. Our findings revealed no significant changes in dendritic arborization, as depicted below (see Author response image 1a).

Author response image 1
Assessment of dendritic arbors in Ta motor pools and proprioceptive innervation in Gl muscles.

(a) Representative dendritic arbors of Ta motor pools retrogradely labeled with rhodamin-dextran (red). Radial plots show similar dendritic development in Isl2 KO. (b) Sensory innervation of muscle spindle of Gl muscles is normal in P28 Isl2 KO, visualized by YFP expression driven by Thy1::YFP.

Our primary focus for synaptic bouton quantification was on Gl motor pools, primarily due to their clear display of ectopic motor neurons apart from those in the typical position. This allowed us to compare the number of synaptic boutons between ectopic and correctly positioned motor neurons. While we suspect that Tfl motor pools might also exhibit similar synaptic defects, these were not analyzed due to their smaller size and challenges associated with locating ectopic neurons. Similarly, other motor pools such as Rf or Ta, which showed relatively normal development, were not subjected to synaptic bouton examination, as the likelihood of synaptic defects in these pools was lower.

References

Arber, S., Ladle, D. R., Lin, J. H., Frank, E., and Jessell, T. M. (2000). ETS gene Er81 controls the formation of functional connections between group Ia sensory afferents and motor neurons. Cell, 101(5), 485-498. doi:10.1016/s0092-8674(00)80859-4

Bonanomi, D., Chivatakarn, O., Bai, G., Abdesselem, H., Lettieri, K., Marquardt, T.,... Pfaff, S. L. (2012). Ret is a multifunctional coreceptor that integrates diffusible- and contact-axon guidance signals. Cell, 148(3), 568-582. doi:10.1016/j.cell.2012.01.024

Catela, C., Chen, Y., Weng, Y., Wen, K., and Kratsios, P. (2022). Control of spinal motor neuron terminal differentiation through sustained Hoxc8 gene activity. ELife, 11. doi:10.7554/eLife.70766

Corchete, L. A., Rojas, E. A., Alonso-Lopez, D., De Las Rivas, J., Gutierrez, N. C., and Burguillo, F. J. (2020). Systematic comparison and assessment of RNA-seq procedures for gene expression quantitative analysis. Sci Rep, 10(1), 19737. doi:10.1038/s41598-020-76881-x

Dasen, J. S., De Camilli, A., Wang, B., Tucker, P. W., and Jessell, T. M. (2008). Hox repertoires for motor neuron diversity and connectivity gated by a single accessory factor, FoxP1. Cell, 134(2), 304-316. doi:10.1016/j.cell.2008.06.019

De Marco Garcia, N. V., and Jessell, T. M. (2008). Early motor neuron pool identity and muscle nerve trajectory defined by postmitotic restrictions in Nkx6.1 activity. Neuron, 57(2), 217-231. doi:10.1016/j.neuron.2007.11.033

El-Brolosy, M. A., and Stainier, D. Y. R. (2017). Genetic compensation: A phenomenon in search of mechanisms. PLoS Genet, 13(7), e1006780. doi:10.1371/journal.pgen.1006780

Evrony, G. D., Lee, E., Park, P. J., and Walsh, C. A. (2016). Resolving rates of mutation in the brain using single-neuron genomics. ELife, 5. doi:10.7554/eLife.12966

Gergics, P., Brinkmeier, M. L., and Camper, S. A. (2015). Lhx4 deficiency: increased cyclin-dependent kinase inhibitor expression and pituitary hypoplasia. Mol Endocrinol, 29(4), 597-612. doi:10.1210/me.2014-1380

Gould, T. W., Yonemura, S., Oppenheim, R. W., Ohmori, S., and Enomoto, H. (2008). The neurotrophic effects of glial cell line-derived neurotrophic factor on spinal motoneurons are restricted to fusimotor subtypes. J Neurosci, 28(9), 2131-2146. doi:10.1523/JNEUROSCI.5185-07.2008

Haase, G., Dessaud, E., Garces, A., de Bovis, B., Birling, M., Filippi, P.,... deLapeyriere, O. (2002). GDNF acts through PEA3 to regulate cell body positioning and muscle innervation of specific motor neuron pools. Neuron, 35(5), 893-905. doi:10.1016/s0896-6273(02)00864-4

Harrison, K. A., Thaler, J., Pfaff, S. L., Gu, H., and Kehrl, J. H. (1999). Pancreas dorsal lobe agenesis and abnormal islets of Langerhans in Hlxb9-deficient mice. Nat Genet, 23(1), 71-75. doi:10.1038/12674

Henderson, C. E., Phillips, H. S., Pollock, R. A., Davies, A. M., Lemeulle, C., Armanini, M.,... et al. (1994). GDNF: a potent survival factor for motoneurons present in peripheral nerve and muscle. Science, 266(5187), 1062-1064. doi:10.1126/science.7973664

Kramer, E. R., Knott, L., Su, F., Dessaud, E., Krull, C. E., Helmbacher, F., and Klein, R. (2006). Cooperation between GDNF/Ret and ephrinA/EphA4 signals for motor-axon pathway selection in the limb. Neuron, 50(1), 35-47. doi:10.1016/j.neuron.2006.02.020

Lee, S. K., Lee, B., Ruiz, E. C., and Pfaff, S. L. (2005). Olig2 and Ngn2 function in opposition to modulate gene expression in motor neuron progenitor cells. Genes Dev, 19(2), 282-294. doi:10.1101/gad.1257105

Livet, J., Sigrist, M., Stroebel, S., De Paola, V., Price, S. R., Henderson, C. E.,... Arber, S. (2002). ETS gene Pea3 controls the central position and terminal arborization of specific motor neuron pools. Neuron, 35(5), 877-892. doi:10.1016/s0896-6273(02)00863-2

Namboori, S. C., Thomas, P., Ames, R., Hawkins, S., Garrett, L. O., Willis, C. R. G.,... Bhinge, A. (2021). Single-cell transcriptomics identifies master regulators of neurodegeneration in SOD1 ALS iPSC-derived motor neurons. Stem Cell Reports, 16(12), 3020-3035. doi:10.1016/j.stemcr.2021.10.010

Sullivan, G. (1962). Anatomy and embryology of the Wing Musculature of the domestic fowl (gallus). Australian Journal of Zoology 10(3), 458-518.

Thaler, J. P., Koo, S. J., Kania, A., Lettieri, K., Andrews, S., Cox, C.,... Pfaff, S. L. (2004). A postmitotic role for Isl-class LIM homeodomain proteins in the assignment of visceral spinal motor neuron identity. Neuron, 41(3), 337-350. doi:10.1016/s0896-6273(04)00011-x

Thaler, J. P., Lee, S. K., Jurata, L. W., Gill, G. N., and Pfaff, S. L. (2002). LIM factor Lhx3 contributes to the specification of motor neuron and interneuron identity through cell-type-specific protein-protein interactions. Cell, 110(2), 237-249. doi:10.1016/s0092-8674(02)00823-1

Vrieseling, E., and Arber, S. (2006). Target-induced transcriptional control of dendritic patterning and connectivity in motor neurons by the ETS gene Pea3. Cell, 127(7), 1439-1452. doi:10.1016/j.cell.2006.10.042

https://doi.org/10.7554/eLife.84596.sa2

Article and author information

Author details

  1. Yunjeong Lee

    School of Life Sciences, Gwangju Institute of Science and Technology, Oryong-dong, Buk-gu, Gwangju, Republic of Korea
    Contribution
    Data curation, Formal analysis, Validation, Investigation
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0009-0002-6694-9991
  2. In Seo Yeo

    School of Life Sciences, Gwangju Institute of Science and Technology, Oryong-dong, Buk-gu, Gwangju, Republic of Korea
    Contribution
    Data curation, Investigation, Methodology
    Competing interests
    No competing interests declared
  3. Namhee Kim

    Fermentation Regulation Technology Research Group, World Institute of Kimchi, Gwangju, Republic of Korea
    Contribution
    Data curation, Formal analysis, Validation, Investigation, Methodology
    Competing interests
    No competing interests declared
  4. Dong-Keun Lee

    School of Life Sciences, Gwangju Institute of Science and Technology, Oryong-dong, Buk-gu, Gwangju, Republic of Korea
    Contribution
    Data curation, Validation, Investigation
    Competing interests
    No competing interests declared
  5. Kyung-Tai Kim

    Jeonbuk Department of Inhalation Research, Korea Institute of Toxicology, Jeongeup-si, Republic of Korea
    Contribution
    Data curation, Validation, Investigation
    Competing interests
    No competing interests declared
  6. Jiyoung Yoon

    School of Life Sciences, Gwangju Institute of Science and Technology, Oryong-dong, Buk-gu, Gwangju, Republic of Korea
    Contribution
    Data curation, Formal analysis, Investigation
    Competing interests
    No competing interests declared
  7. Jawoon Yi

    School of Life Sciences, Gwangju Institute of Science and Technology, Oryong-dong, Buk-gu, Gwangju, Republic of Korea
    Contribution
    Formal analysis
    Competing interests
    No competing interests declared
  8. Young Bin Hong

    1. Department of Neurology, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, Republic of Korea
    2. Department of Biochemistry, College of Medicine, Dong-A University, Busan, Republic of Korea
    Contribution
    Data curation, Validation, Investigation, Methodology
    Competing interests
    No competing interests declared
  9. Byung-Ok Choi

    Department of Neurology, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, Republic of Korea
    Contribution
    Resources, Supervision
    Competing interests
    No competing interests declared
  10. Yoichi Kosodo

    Korea Brain Research Institute, Daegu, Republic of Korea
    Contribution
    Methodology
    Competing interests
    No competing interests declared
  11. Daesoo Kim

    Department of Brain and Cognitive Sciences, Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Republic of Korea
    Contribution
    Methodology
    Competing interests
    No competing interests declared
  12. Jihwan Park

    School of Life Sciences, Gwangju Institute of Science and Technology, Oryong-dong, Buk-gu, Gwangju, Republic of Korea
    Contribution
    Data curation, Formal analysis, Validation, Investigation
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5728-912X
  13. Mi-Ryoung Song

    School of Life Sciences, Gwangju Institute of Science and Technology, Oryong-dong, Buk-gu, Gwangju, Republic of Korea
    Contribution
    Conceptualization, Supervision, Funding acquisition, Writing - original draft
    For correspondence
    msong@gist.ac.kr
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-0350-0863

Funding

National Research Foundation of Korea (NRF-2018R1A5A1024261)

  • Mi-Ryoung Song

National Research Foundation of Korea (NRF-2022M3E5E8081194)

  • Mi-Ryoung Song

National Research Foundation of Korea (NRF-2023R1A2C1002690)

  • Mi-Ryoung Song

Gwangju Institute of Science and Technology (the GIST Research Institute (GRI) IIBR in 2023)

  • Mi-Ryoung Song

SNUH Lee Kun-hee Child Cancer & Rare Disease Project (22B-001-0500)

  • Mi-Ryoung Song

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

This work was supported by grants from NRF (NRF-2018R1A5A1024261, NRF-2022M3E5E8081194, NRF-2023R1A2C1002690), SNUH Lee Kun-hee Child Cancer & Rare Disease Project (22B-001-0500) and a GIST Research Institute (GRI) IIBR funded by the GIST in 2023. We thank Drs. Samuel Pfaff, Jeremy Dasen, and Bennett Novitch for their generous contribution of the research materials used in the study. We thank Jung Eun Kim and Won-Young Lee for their support in experimental setup, Hwan Kim (GIST Central Research Facility) for technical assistance in microscopy, and Myungin Baek for critical comments on the manuscript.

Ethics

This study was performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of Gwangju Institute of Science and Technology (GIST, No. GIST2021-072). All mice were housed and cared for in an Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC)-accredited animal facility under specific pathogen-free conditions.

Senior Editor

  1. Marianne E Bronner, California Institute of Technology, United States

Reviewing Editor

  1. Paschalis Kratsios, University of Chicago, United States

Reviewers

  1. Paschalis Kratsios, University of Chicago, United States
  2. Niccolò Zampieri, Max Delbrück Center for Molecular Medicine, Germany

Version history

  1. Received: October 31, 2022
  2. Preprint posted: November 8, 2022 (view preprint)
  3. Accepted: October 11, 2023
  4. Accepted Manuscript published: October 23, 2023 (version 1)
  5. Version of Record published: November 10, 2023 (version 2)

Copyright

© 2023, Lee et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 374
    Page views
  • 80
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Yunjeong Lee
  2. In Seo Yeo
  3. Namhee Kim
  4. Dong-Keun Lee
  5. Kyung-Tai Kim
  6. Jiyoung Yoon
  7. Jawoon Yi
  8. Young Bin Hong
  9. Byung-Ok Choi
  10. Yoichi Kosodo
  11. Daesoo Kim
  12. Jihwan Park
  13. Mi-Ryoung Song
(2023)
Transcriptional control of motor pool formation and motor circuit connectivity by the LIM-HD protein Isl2
eLife 12:e84596.
https://doi.org/10.7554/eLife.84596

Further reading

    1. Chromosomes and Gene Expression
    2. Developmental Biology
    Virginia L Pimmett, Mounia Lagha
    Insight

    Imaging experiments reveal the complex and dynamic nature of the transcriptional hubs associated with Notch signaling.

    1. Cell Biology
    2. Developmental Biology
    Simon Schneider, Andjela Kovacevic ... Hubert Schorle
    Research Article

    Cylicins are testis-specific proteins, which are exclusively expressed during spermiogenesis. In mice and humans, two Cylicins, the gonosomal X-linked Cylicin 1 (Cylc1/CYLC1) and the autosomal Cylicin 2 (Cylc2/CYLC2) genes, have been identified. Cylicins are cytoskeletal proteins with an overall positive charge due to lysine-rich repeats. While Cylicins have been localized in the acrosomal region of round spermatids, they resemble a major component of the calyx within the perinuclear theca at the posterior part of mature sperm nuclei. However, the role of Cylicins during spermiogenesis has not yet been investigated. Here, we applied CRISPR/Cas9-mediated gene editing in zygotes to establish Cylc1- and Cylc2-deficient mouse lines as a model to study the function of these proteins. Cylc1 deficiency resulted in male subfertility, whereas Cylc2-/-, Cylc1-/yCylc2+/-, and Cylc1-/yCylc2-/- males were infertile. Phenotypical characterization revealed that loss of Cylicins prevents proper calyx assembly during spermiogenesis. This results in decreased epididymal sperm counts, impaired shedding of excess cytoplasm, and severe structural malformations, ultimately resulting in impaired sperm motility. Furthermore, exome sequencing identified an infertile man with a hemizygous variant in CYLC1 and a heterozygous variant in CYLC2, displaying morphological abnormalities of the sperm including the absence of the acrosome. Thus, our study highlights the relevance and importance of Cylicins for spermiogenic remodeling and male fertility in human and mouse, and provides the basis for further studies on unraveling the complex molecular interactions between perinuclear theca proteins required during spermiogenesis.