Ectopic expression of Irx3 and Irx5 in the paraventricular nucleus of the hypothalamus contributes to defects in Sim1 haploinsufficiency.

The paraventricular nucleus of the hypothalamus (PVH) contains a heterogeneous cluster of Sim1-expressing neurons critical for feeding regulation. Sim1 haploinsufficiency results in hyperphagic obesity with disruption of PVH neurons, yet the molecular profiles of PVH neurons and the mechanism underlying the defects of Sim1 haploinsufficiency are not well understood. By single-cell RNA sequencing, we identified two major populations of Sim1+ PVH neurons, which are differentially affected by Sim1 haploinsufficiency. The Iroquois homeobox genes Irx3 and Irx5 have been implicated in the hypothalamic control of energy homeostasis. We found that Irx3 and Irx5 are ectopically expressed in the Sim1+ PVH cells of Sim1+/− mice. By reducing their dosage and PVH-specific deletion of Irx3, we demonstrate that misexpression of Irx3 and Irx5 contributes to the defects of Sim1+/− mice. Our results illustrate abnormal hypothalamic activities of Irx3 and Irx5 as a central mechanism disrupting PVH development and feeding regulation in Sim1 haploinsufficiency.

INTRODUCTION

The paraventricular nucleus of the hypothalamus (PVH) is a complex brain region consisting of a highly heterogeneous population of neurons with diverse physiological functions including appetite regulation (–). In the mouse, PVH neurons are generated between embryonic day 10.5 (E10.5) and E12.5, and their terminal differentiation completes largely before birth (–). Mutant studies in mice have illustrated that the specification of PVH neurons is programmed by a cascade of transcription factors (TFs) including single-minded 1 (SIM1), which is highly expressed in the anterior hypothalamus during embryogenesis and persists in most PVH neurons of adult brain (, , ). Using SIM1 as a specific marker of PVH neurons, the functions of various neuronal subtypes including peptidergic neurons such as vasopressin (AVP+), oxytocin (OXT+), corticotropin-releasing hormone (CRH+), thyrotropin-releasing hormone (TRH+), and prodynorphin (PDYN+) neurons, as well as nonpeptidergic neurons such as melanocortin-4 receptor (MC4R+) and nitric oxide synthase–1 (NOS1+) neurons, have been examined in appetite regulation (, –). However, although SIM1+ PVH neurons have been characterized and are known to be heterogeneous, their molecular identities remain poorly understood at the single-cell level.

In humans, loss-of-function mutations in SIM1 have been implicated in hyperphagic childhood obesity (–). Knockout (KO) of Sim1 in mice (Sim1, Sim1KO) results in perinatal lethality and the absence of multiple neuroendocrine cell types in the PVH, unveiling a pivotal role for SIM1 in the formation of PVH neurons (). Similar to humans, early onset of hyperphagic obesity was found in mice with heterozygous loss of Sim1 (, ). Sim1Het (Sim1) mice exhibit a major reduction of AVP+ and OXT+ neurons, but not CRH+ neurons, in the PVH and compromised melanocortin signaling (–), suggesting that Sim1 haploinsufficiency results in specific neurodevelopmental defects in the PVH leading to the hyperphagia phenotype. Furthermore, postnatal knockdown or KO of Sim1 in the PVH neurons of adult mice also resulted in hyperphagia, indicating that Sim1 function is required in mature PVH neurons for appetite regulation (–).

The Iroquois homeobox genes IRX3 and IRX5 have been implicated as determinants of human obesity in connection with the intronic variants of FTO (–). We have recently demonstrated that Irx3 and Irx5 are expressed in multiple cell types of the mouse hypothalamic arcuate-median eminence (ARC-ME), and their activities regulate hypothalamic leptin sensitivity and food intake (). Irx3 was previously shown to be ectopically expressed in the anterior hypothalamus of E12.5 Sim1KO embryos (), suggesting that this misexpression has the potential to disrupt the formation of PVH neurons in Sim1KO mice. Thus, we are intrigued to investigate whether ectopic misexpression of Irx3 and Irx5 is also found in the anterior hypothalamus of Sim1Het mice, and whether they contribute to the neurodevelopmental defects of the PVH underlying the hyperphagia phenotype of Sim1 haploinsufficiency.

In this study, we used the Sim1-Cre;Ai14 tdTomato labeling system (, ) to track the Sim1+ cell lineage along with single-cell RNA sequencing (scRNA-seq) analysis to explore the cellular and molecular profiles of the PVH neurons and identified two major populations of Sim1-Cre+ PVH neurons, which are differentially affected by Sim1 haploinsufficiency. We demonstrated that Irx3 and Irx5 are ectopically expressed in the Sim1-Cre+ PVH cells of Sim1Het mice. Through reduction of Irx3 or/and Irx5 dosage as well as conditional KO of Irx3 in Sim1-Cre+ PVH cells, our analysis demonstrates that the abnormal expression of Irx3 and Irx5 leads to the neurodevelopmental defects and hyperphagia in Sim1 haploinsufficiency. These observations establish a novel disease mechanism underlying the hyperphagic obesity of Sim1 haploinsufficiency and illustrate that Irx3 and Irx5 are also involved in obesity due to defective feeding regulation.

RESULTS

scRNA-seq analysis reveals two major PVH neuronal populations

To define the cellular and molecular profiles of PVH neurons, as well as the effects of Sim1 haploinsufficiency on PVH neurons, we performed scRNA-seq analysis of Sim1-Cre+ hypothalamic cells from control and Sim1Het neonatal mice. We used Sim1-Cre () and the Ai14 tdTomato reporter system () to mark PVH cells at postnatal day 1 (P1) to P4, when differentiation of PVH neurons is complete. Through fluorescence-activated cell sorting, we isolated tdTomato+ (Sim1-Cre+) cells from hypothalami excluding the supraoptic nucleus, which also contains Sim1+ cells, and prepared single-cell libraries for transcriptome analysis (Fig. 1A).

Fig. 1.

scRNA-seq analysis of Sim1-Cre+ cells reveal characteristics of PVH neurons and their disruption in Sim1 haploinsufficiency.

(A) Schematic illustration of live (Sytox blue−) tdTomato+ cells prepared from hypothalami of P1-P4 Sim1-Cre;Ai14 tdTomato mice for scRNA-seq analysis. DAPI, 4′,6-diamidino-2-phenylindole. (B) UMAP plot of sequenced tdTomato+ cells annotated according to known cell type markers. (C) Heatmap of top marker genes for each cell cluster and annotation. (D) Dot plot showing the expression levels (color) and percentages (size) of cells expressing TFs and neuropeptides/enzymes enriched in each cluster. (E) Proportional distribution of cell types as revealed by scRNA-seq analysis. (F) Cell type proportional changes in Sim1Het versus control samples based on marker gene expression. (G) Representative images and (H) quantification of the number of AVP+ cells, NOS1+ cells, OXT+ cells, and CART+ cells in the PVH of P0 mice (n = 7/7 for AVP+; n = 6/6 for NOS1+; n = 7/7 for OXT+; n = 8/5 for CART+). Data are represented as means ± SEM values. *P < 0.05; **P < 0.01. WT, wild type. Scale bars, 100 μm.

Through graph-based clustering, we projected 5283 individual cells (2807 from control and 2476 from Sim1Het) onto a uniform manifold approximation and projection (UMAP) plot, yielding three distinct cell clusters (Fig. 1B). Using the expression patterns of previously established marker genes and differentially expressed genes found in each cell cluster, we identified a progenitor population enriched in Sox2 expression and two neuronal populations (“neuron 1” and “neuron 2”), both with high expression of pan-neuronal markers Elavl3 and Tubb3 (Fig. 1, B to D, and fig. S1). Neuron 1 is enriched in the expression of genes encoding TFs including Foxb1, Lhx1, and Pou2f2, as well as the Cck, Cartpt, and Nts neuropeptides, whereas neuron 2 shows high gene expression of TFs, including Otp, Nr2f2 (COUP-TFII), Pou3f2 (Brn2), and Onecut2, and neuropeptides/enzymes, including Avp, Oxt, Pdyn, Trh, Crh, Nos1, and Mc4r (Fig. 1D and fig. S1). Immunostaining of markers for the two neuronal populations [Cocaine- and amphetamine-regulated transcript (CART) for neuron 1 and AVP, NOS1, and OXT for neuron 2] confirms the presence of these neuronal subtypes in the PVH (Fig. 1G and fig. S3C). It is important to note that Orthopedia (OTP) and POU domain, class 3, transcription factor 2 (POU3F2), which are highly expressed in neuron 2, are known to act in concert with SIM1 in the terminal differentiation of specific PVH neurons, e.g., AVP+ and OXT+ neurons (, , , –). Neuron 2 contains many known players in feeding regulation including AVP+, NOS1+, OXT+, PDYN+, and MC4R+ neurons (, –). In contrast, the functional role of neuron 1 cells in the control of energy homeostasis is not well understood. These results unveil the presence of two major molecularly distinct Sim1-Cre+ neuronal populations in the PVH, further illustrating the heterogeneity of PVH neurons.

Sim1 haploinsufficiency disrupts formation of specific PVH neurons

Sim1Het mice display a reduction in PVH cell density, area, and number of nuclear profiles (, ). However, analysis of Sim1Het mice using a Sim1–green fluorescent protein (GFP) reporter did not reveal a significant difference in the number and area of GFP-positive neurons in the PVH (). By Sim1-Cre;Ai14 tdTomato labeling of the PVH, we also did not observe any difference between control and Sim1Het mice in the number of tdTomato+ cells or the proportion of neurons as marked by HuC (fig. S2, A to C). Consistent with this, similar proportions of neurons were detected in the Sim1-Cre+ scRNA-seq datasets of control and Sim1Het mice (fig. S2, D to F). Despite the observation that overt neuronal differentiation is not affected by Sim1 haploinsufficiency, we observed a marked difference in the composition of neuronal populations in Sim1 mice. When compared with the control, there is a significant increase of the neuron 1 population accompanied by a reduction of the neuron 2 population (Fig. 1E). The proportion of cells expressing markers enriched in neuron 1 was higher in Sim1Het mice, whereas that of cells expressing markers enriched in neuron 2 was lower in Sim1Het mice (Fig. 1F and fig. S3, A and B). Specifically, formation of some neuron 2 cell types including AVP+ and OXT+ neurons, as well as NOS1+ neurons, is negatively affected by Sim1 haploinsufficiency, whereas other neuron 2 cell types such as CRH+ and TRH+ neurons are not as sensitive to Sim1 gene dosage. Our data unveiled that a subset of PVH neurons (neuron 1) is augmented by Sim1 haploinsufficiency. Consistent with previous results (, ), we showed by immunostaining that there is a marked reduction in the number of AVP+ and OXT+ PVH neurons (neuron 2) in both P0 and adult Sim1Het mice (Fig. 1, G and H, and fig. S3, C and D). In addition, we demonstrated a reduced number of NOS1+ neurons (neuron 2) and an increased number of CART+ neurons (neuron 1) in the PVH of Sim1Het mice. This analysis indicates that Sim1 haploinsufficiency affects the formation of a specific subset of PVH neurons.

Irx3 and Irx5 are ectopically expressed in Sim1-Cre+ PVH cells of Sim1Het mice

In E11.5 Sim1KO mouse embryos (), Irx3 was found to be ectopically expressed in the anterior hypothalamus, which gives rise to the PVH. To investigate whether Sim1Het mice show ectopic expression of Irx3 and Irx5, we performed RNA in situ hybridization and quantitative polymerase chain reaction (qPCR) analyses. Intriguingly, we found that Irx3 and Irx5 are both misexpressed in the anterior hypothalamus of E11.5 Sim1Het embryos, and their ectopic expression persists in the PVH of Sim1Het mice at P0 (Fig. 2, A and B). In contrast, we did not detect any abnormal expression of Irx3 and Irx5 in the hypothalamic ARC-ME of Sim1Het mice (Fig. 2, A and C). By costaining tdTomato and Irx3 or Irx5 RNA, we confirmed that Irx3 and Irx5 are misexpressed in Sim1-Cre+ PVH cells of Sim1Het mice (Fig. 2D). This was further supported by qPCR analysis of tdTomato-positive cells sorted from the hypothalamus of Sim1-Cre;Ai14 tdTomato mice (Fig. 2E). However, we were unable to detect ectopic expression of Irx3 and Irx5 in the Sim1-Cre+ cells from Sim1Het mice by scRNA-seq analysis. We postulate that this is due to the limited sequencing depth of the droplet-based scRNA-seq pipeline (, ). Together, our results illustrate the ectopic expression of Irx3 and Irx5 in the Sim1-Cre+ PVH cells of Sim1Het mice.

Fig. 2.

Ectopic expression of Irx3 and Irx5 in the Sim1-Cre+ PVH cells of Sim1Het mice.

(A) Histological maps illustrating the expression pattern of Irx3 and Irx5 in the WT and Sim1Het hypothalamus. Section (i) for the PVH and section (ii) for the ARC-ME. 3V, third ventricle. (B) In situ hybridization of Irx3 and Irx5 mRNA in the anterior hypothalamus (black arrows) of E11.5 embryos and PVH (black arrows) of P0 mice. (C) In situ hybridization of Irx3 and Irx5 mRNA in the ARC-ME (dashed lines) of P0 and adult mice. (D) Fluorescence in situ hybridization of Irx3, Irx5, and tdTomato RNA in the PVH of P0 Sim1-Cre;Ai14 tdTomato and Sim1-Cre;Ai14 tdTomato;Sim1Het mice. (E) Top: Isolation of Sim1-Cre+ cells from P0 hypothalami using the Sim1-Cre;Ai14 tdTomato system. Bottom: Gene expression analysis in Sim1-Cre+ cells from control and Sim1Het mice (Cont/Sim1Het: n = 3/3). Data are represented as means ± SEM values. **P < 0.01. Scale bars, 100 μm.

A half-reduction of Irx3 or Irx5 dosage in Sim1Het mice suppresses hyperphagic obesity

To test whether abnormal expression of Irx3 and Irx5 in the PVH of Sim1Het mice contributes to the hyperphagic obesity phenotype, we generated Sim1Het mouse lines lacking one copy of Irx3 (Sim1Het;Irx3Het) or Irx5 (Sim1Het;Irx5Het) and confirmed that expression levels of Irx3 or Irx5 in Sim1-Cre+ PVH cells are reduced by ~50% after deleting one allele of Irx3 or Irx5, respectively, without affecting Sim1 expression in these mutants (figs. S4A and S5A). We further examined the effects of a half-reduction of Irx3 or Irx5 dosage on hyperphagia and other metabolic phenotypes. While Irx3Het and Irx5Het mice appeared wild type and did not show any obvious body weight and mass phenotypes (), both Sim1Het;Irx3Het and Sim1Het;Irx5Het mice exhibited a less obese phenotype including a clear reduction of body weight, adiposity, and other metabolic abnormalities (i.e., hepatosteatosis and glucose intolerance) compared to Sim1Het mice (Fig. 3, A to F, and I to N). Notably, reduction of Irx3 or Irx5 dosage significantly suppressed excessive food intake without affecting energy expenditure, respiratory exchange ratio (RER), and locomotor activity (Fig. 3, G, H, and O to P, and figs. S4, B to F, and S5, B to F). These results support the notion that misexpression of Irx3 and Irx5 in Sim1Het mice is involved in the hyperphagia phenotype.

Fig. 3.

Reduced dosage of Irx3 or Irx5 in Sim1Het mice partially suppresses hyperphagic obesity.

(A) Gross appearance of WT, Sim1Het, and Sim1Het;Irx3Het mice at 11 weeks old. Photo credit: Joe Eun Son, The Hospital for Sick Children. (B) Body weight and the inset bar graph show body weight of 11-week-old mice (WT, Sim1Het, and Sim1Het;Irx3Het; n = 13/11/10). (C) Body composition analysis of 12-week-old mice (n = 13/11/8). (D) Tissue weight of perigonadal white adipose tissue (PWAT), inguinal WAT (IWAT), brown adipose tissue (BAT), and liver (n = 13/13/11). (E) Glucose tolerance tests (GTTs). The inset bar graph shows area under curve (AUC). n = 9/7/6. (F) Hematoxylin and eosin (H&E) staining of PWAT, BAT, and liver sections. (G) Daily energy expenditure adjusted with lean mass of 29.0 g using analysis of covariance (ANCOVA) of WT, Sim1Het, and Sim1Het;Irx3Het mice (n = 12/10/12). (H) Food intake. The inset bar graph shows total food intake until 11 weeks of age (n = 9/6/6). (I) Gross appearance of WT, Sim1Het, and Sim1Het;Irx5Het mice at 11 weeks old. Photo credit: Joe Eun Son, The Hospital for Sick Children. (J) Body weight. The inset bar graph shows body weight of 11-week-old mice (WT, Sim1Het, and Sim1Het;Irx5Het; n = 15/11/21). (K) Body composition analysis at 12 weeks of age (n = 9/8/9). (L) Tissue weight of PWAT, IWAT, BAT, and liver (n = 6/8/13). (M) GTTs. The inset bar graph shows AUC. n = 9/7/10. (N) H&E staining of PWAT, BAT, and liver sections. (O) Daily energy expenditure adjusted with a lean mass of 29.6 g using ANCOVA of WT, Sim1Het, and Sim1Het;Irx5Het mice (n = 9/5/6). (P) Food intake. The inset bar graph shows total food intake until 11 weeks of age (n = 7/6/7). Data are represented as means ± SEM values. *P < 0.05; **P < 0.01, ns, not significant. Scale bars, 100 μm.

A simultaneous reduction of Irx3 and Irx5 completely overrides obesity in Sim1Het mice

Because a half-reduction of Irx3 or Irx5 dosage partially suppressed the phenotype of Sim1Het mice, we investigated whether simultaneous reduction of Irx3 and Irx5 can further ameliorate the hyperphagic obesity due to Sim1 haploinsufficiency. By using mice harboring the cis-heterozygous mutant alleles of Irx3 and Irx5 (Irx3/5dHet), we generated Sim1Het;Irx3/5dHet mice and confirmed a half-reduction of Irx3 and Irx5 expression, but not Sim1, in Sim1-Cre+ PVH cells of Sim1Het;Irx3/5dHet mice compared to Sim1Het mice (fig. S6A). In our recent study (), simultaneous reduction of Irx3 and Irx5 dosage in mice (Irx3/5dHet) was shown to result in a lean phenotype through increased energy expenditure due to adipose browning and elevated leptin sensitivity with reduced food intake. Notably, Sim1Het;Irx3/5dHet mice exhibited a lean phenotype almost identical to that of Irx3/5dHet mice (Fig. 4, A and B). Both Sim1Het;Irx3/5dHet and Irx3/5dHet mice were leaner than wild-type mice and showed reduced adiposity (Fig. 4, C to E). Furthermore, major metabolic complications of obesity such as hepatosteatosis and glucose intolerance in Sim1Het mice were completely rescued by the simultaneous reduction of Irx3 and Irx5 dosage (Fig. 4, E and F). Similar to Irx3/5dHet mice, Sim1Het;Irx3/5dHet mice also showed increased energy expenditure and lower food intake (Fig. 4, G and H, and fig. S6B) but without observable differences in RER and locomotor activity (fig. S6, C to F). These results reveal that the haploinsufficiency effects of Sim1 on food intake regulation and obesity can be completely overridden in mice by a simultaneous half-reduction of Irx3 and Irx5 dosage, pinpointing the central roles of Irx3 and Irx5 activities underlying the hyperphagic obesity of Sim1Het mice.

Fig. 4.

A simultaneous reduction of Irx3 and Irx5 in Sim1Het mice overrides hyperphagia.

(A) Gross appearance of 11-week-old mice. Photo credit: Joe Eun Son, The Hospital for Sick Children. (B) Body weight of WT, Sim1Het, Sim1Het;Irx3/5dHet, and Irx3/5dHet mice. The inset bar graph shows body weight of 11-week-old mice (n = 18/13/17/21). (C) Body composition analysis of WT, Sim1Het, Sim1Het;Irx3/5dHet, and Irx3/5dHet mice at 12 weeks of age (n = 11/13/8/13). (D) Tissue weight of PWAT, IWAT, BAT, and liver (n = 13/12/11/14). (E) H&E staining of PWAT, BAT, and liver sections. (F) GTTs. The inset bar graph shows AUC (n = 10/12/17/11). (G) Daily energy expenditure adjusted with a lean mass of 26.1 g using ANCOVA (n = 10/11/10/9). (H) Food intake. The inset bar graph shows total food intake until 11 weeks of age (n = 8/10/12/14). Data are represented as means ± SEM values. *P < 0.05; **P < 0.01. Scale bars, 100 μm.

Reduction of Irx3 and Irx5 restores formation of PVH neurons in Sim1Het mice

Next, we sought to determine whether reduction of Irx3 and Irx5 in Sim1Het mice can restore the PVH neuronal populations to suppress hyperphagia. Notably, we found that simultaneous reduction of Irx3 and Irx5 dosage significantly restores PVH neuronal populations in Sim1Het mice. The number of AVP+, NOS1+, and OXT+ PVH neurons (neuron 1) was significantly restored in Sim1Het;Irx3/5dHet mice compared to Sim1Het mice (Fig. 5, A and B). Furthermore, the number of CART+ PVH neurons (neuron 2) was also significantly reduced in Sim1Het;Irx3/5dHet mice to a level closer to wild-type mice. Consistent with this, activation of these PVH neurons by the melanocortin agonist melanotan-2 (MTII), as revealed by staining of c-Fos+ cells (Fig. 5, C and D), was markedly enhanced in Sim1Het;Irx3/5dHet mice compared with Sim1Het mice, suggesting that melanocortin signaling is restored by reduction of Irx3 and Irx5 in the PVH of Sim1Het mice. While these phenotypes are only partially rescued when compared with wild-type controls and Irx3/5dHet mice, these results indicate that misexpression of Irx3 and Irx5 indeed contributes to the neurodevelopmental defects of Sim1Het mice.

Fig. 5.

Reduction of Irx3 and Irx5 in Sim1Het mice restores the PVH neuronal population.

(A) Representative images and (B) quantification of the number of AVP+ cells, NOS1+ cells, OXT+ cells, and CART+ cells in the PVH (n = 6/7/6/7 for AVP+; n = 6/6/7/7 for NOS1+; n = 6/7/7/10 for OXT+; n = 6/6/7/7 for CART+). (C) Representative images and (D) quantification of the number of c-Fos+ cells in the PVH following MTII injection (vehicle: n = 7/7/7/7; MTII: n = 8/8/7/6). Data are represented as means ± SEM values. *P < 0.05; **P < 0.01. Scale bars, 100 μm.

PVH deletion of Irx3 in Sim1Het mice restores the hyperphagic obesity phenotype and neuronal disruption

Next, to specifically examine the effects of ectopic Irx3 expression in the Sim1-Cre+ PVH cells on the hyperphagic and neurodevelopmental phenotype in Sim1Het mice, we generated mice with deletion of Irx3 in the Sim1+ lineage using an Irx3-floxed allele (Irx3Fl) () and the Sim1-Cre mouse line () (Fig. 6A). Sim1-Cre;Irx3Fl/Fl (hereafter PVH-Irx3KO) mice did not exhibit any observable differences in body weight, adiposity, and PVH neuronal populations (fig. S7), consistent with the lack of Irx3 and Irx5 expression in normal PVH (Fig. 2, A and B). These results suggest that Irx3, and likely Irx5, is dispensable in the normal PVH. On the other hand, when Irx3 is specifically deleted in the Sim1-Cre+ PVH cells of Sim1Het mice (Sim1Het;Sim1-Cre;Irx3Fl/+ and Sim1Het;Sim1-Cre;Irx3Fl/Fl, hereafter Sim1Het;PVH-Irx3Het and Sim1Het;PVH-Irx3KO, respectively) (Fig. 6A and fig. S8A), the body weight and adiposity of Sim1Het mice were markedly reduced in an Irx3 dosage-dependent manner (Fig. 6, B to D and F). Further analysis of Sim1Het;PVH-Irx3KO mice revealed that obesity-driven metabolic abnormalities such as glucose intolerance and hepatosteatosis are markedly suppressed by Sim1-Cre mediated deletion of Irx3 in Sim1Het mice (Fig. 6, E and F). Furthermore, Sim1Het;PVH-Irx3KO mice exhibited no significant differences in energy expenditure, RER, and locomotor activity when compared with control and Sim1Het mice (Fig. 6G and fig. S8, B to F) but showed a major reduction in food intake (Fig. 6H), indicating that the reduced body weight of these mice is due to the suppression of the hyperphagia phenotype of Sim1 haploinsufficiency. Similar to those observed in Sim1Het;Irx3/5dHet mice, the number of AVP+, NOS1+, and OXT+ PVH neurons as well as CART+ PVH neurons were largely restored to a control level in Sim1Het;PVH-Irx3KO mice (Fig. 6, I and J). As revealed by staining of c-Fos+ cells, activation of PVH neurons by the melanocortin agonist MTII was clearly elevated in Sim1Het;PVH-Irx3 compared to Sim1Het mice (Fig. 6, K and L). These observations indicate that deletion of Irx3 in the PVH of Sim1Het mice could significantly restore the PVH neuronal population and their response to melanocortin. It is noteworthy that Sim1Het;PVH-Irx3KO mice still exhibit mild hyperphagic obesity and PVH neuronal disruption phenotypes compared to control mice (Fig. 6), which may be due to remaining Irx5 misexpression in the PVH of these mice (fig. S8A). Together, our data support the notion that misexpression of Irx3 and likely Irx5 in the Sim1-Cre+ PVH cells is largely responsible for the neurodevelopmental and hyperphagic obesity phenotypes of Sim1Het mice.

Fig. 6.

Deletion of Irx3 in Sim1-Cre+ cells of Sim1Het mice suppresses hyperphagic obesity and restores PVH neuron formation.

(A) Breeding scheme for the generation of Sim1Het;Sim1-Cre;Irx3Fl/Fl (Sim1Het;PVH-Irx3KO) mice. (B) Body weight of control, Sim1Het, Sim1Het;PVH-Irx3Het, and Sim1Het;PVH-Irx3KO mice. The inset bar graph shows body weight of 11-week-old mice (n = 24/11/15/16). (C) Body composition analysis (n = 15 Cont, n = 12 Sim1Het, n = 9 Sim1Het;PVH-Irx3Het, n = 15 Sim1Het;PVH-Irx3KO). (D) Tissue weight of PWAT, IWAT, BAT, and liver (n = 15/12/9/14). (E) GTTs. The inset bar graph shows AUC (n = 13/10/6). (F) H&E staining of PWAT, BAT, and liver sections. (G) Daily energy expenditure adjusted with lean mass of 27.6 g using ANCOVA (n = 11/12/8). (H) Food intake. The inset bar graph shows accumulated food intake until 11 weeks of age (n = 9/9/13). (I) Representative images and (J) quantification of the number of AVP+ cells, NOS1+ cells, OXT+ cells, and CART+ cells in the PVH (n = 8/9/10 for AVP+ or NOS1+ or OXT+; n = 6/7/8 for CART+). (K) Representative images and (L) quantification of the number of c-Fos+ cells in the PVH following MTII injection (vehicle: n = 4/3/4; MTII: n = 5/4/5). Data are represented as means ± SEM values. *P < 0.05; **P < 0.01. Scale bars, 100 μm.

DISCUSSION

In this study, we uncovered ectopic PVH expression of Irx3 and Irx5 as a novel disease mechanism underlying the hyperphagic obesity of Sim1 haploinsufficiency in mice (Fig. 7). These results suggest that IRX3 and IRX5 are also involved in monogenic obesity disorders of SIM1 haploinsufficiency (–) besides their implicated roles in FTO-associated obesity at the population level (, ). It has been established that defective feeding regulation leading to hyperphagia is caused by insufficient Sim1 activity, as CRISPR-mediated activation of the remaining functional Sim1 gene in Sim1Het mice could rescue the obesity phenotype (). However, the molecular mechanisms underlying Sim1 haploinsufficiency remain undefined. By using various genetic mouse models, our analysis here provides strong evidence demonstrating that the neurodevelopmental defects in Sim1Het mice are at least partly due to ectopic expression of Irx3 and Irx5 in the PVH.

Fig. 7.

Model depicting abnormal expression of Irx3 and Irx5 in the PVH leads to neurodevelopmental defects and hyperphagic obesity in Sim1 haploinsufficiency.

Two major subpopulations (neuron 1 and neuron 2) of Sim1-Cre+ PVH neurons are differentially affected by Sim1 haploinsufficiency. Irx3 and Irx5 are ectopically expressed in the Sim1-Cre+ PVH cells of Sim1Het mice and contribute to the PVH neuronal disruption and hyperphagic obesity of Sim1 haploinsufficiency. Reduction of Irx3 or/and Irx5 dosage in Sim1Het mice restores the formation of PVH neurons rescuing the hyperphagic obesity phenotype.

Recently, we demonstrated that Irx3 and Irx5 play overlapping functions in the control of food intake and energy expenditure (). Irx3/5dHet mice exhibit adipose browning resulting in elevated energy expenditure. In addition, these mice show higher hypothalamic leptin sensitivity in the ARC-ME with a lower food intake. Together, these metabolic features lead to an overall whole-body phenotype, which is leaner than wild-type mice. Notably, we found that a half-reduction of Irx3 and Irx5 dosage can override the haploinsufficient effects of Sim1 on hyperphagia, resulting in a lean phenotype in Sim1Het;Irx3/5dHet mice similar to that of Irx3/5dHet mice (Fig. 4). In particular, the food intake of Sim1Het;Irx3/5dHet mice was lower than that of wild-type mice (Fig. 4) despite the incomplete rescue of PVH neuronal defects in Sim1Het;Irx3/5dHet mice (Fig. 5). We postulate that besides a restoration of the feeding regulatory neuronal circuitry in the PVH, lower food intake is likely driven by higher leptin sensitivity in the hypothalamic ARC-ME region of these mice due to their Irx3/5dHet background. Moreover, elevated energy expenditure in the Irx3/5dHet background also contributes to the lean phenotype. The unexpected observation that the effects of Sim1 haploinsufficiency can be completely nullified by a half-reduction of Irx3 and Irx5 dosage in mice further illustrates the importance of IRX3 and IRX5 in energy homeostasis and human obesity.

Several lines of research have suggested regulatory functions of Irx genes in the development of the nervous system. Members of the Drosophila Irx gene family are implicated in the development of the peripheral nervous system, and Xenopus Irx genes are involved in neurogenesis (–). In mice, Irx genes are also expressed during early neurogenesis and implicated in the regulation of Ascl1 (Mash1). In the developing nervous system, Irx3 and Irx5 display an overlapping expression pattern with Ascl1 (). Notably, Ascl1 is critical for the differentiation and specification of hypothalamic neuroendocrine cells (, ), suggesting potential involvement of Irx3 and Irx5 in neuronal development and specification. Our recent study has demonstrated that Irx3 and Irx5 expression in the radial glia-like neural stem cells regulates postnatal neurogenesis in the hypothalamic ARC (). Here, we show that abnormal expression of Irx3 and Irx5 in Sim1+ PVH cells in Sim1 haploinsufficiency contributes to neurodevelopmental defects via disruption of the neuronal populations involved in feeding regulation. Future studies are warranted to investigate the molecular mechanisms underlying Irx3- and Irx5-mediated PVH neuronal defects in Sim1 haploinsufficiency. To delineate this question, deletion of both Irx3 and Irx5 in Sim1+ cells would be an ideal experiment. Defining the genes/pathways activated or repressed by Irx3 and Irx5 in Sim1+ PVH cells should yield deeper insights about neuronal development and specification of PVH.

Many Sim1+ PVH neuron subtypes have been established for their roles in feeding regulation (, –, ). In this study, our lineage tracing and scRNA-seq analyses have revealed additional heterogeneity of Sim1-Cre+ PVH neurons. Furthermore, we showed that Sim1 haploinsufficiency affects the formation of specific PVH neurons but not overt neuronal differentiation. On the basis of TF and neuropeptide gene signatures, Sim1-Cre+ PVH neurons can be clustered into two main groups (Fig. 1D). The neuron 2 population consists of ~80% of Sim1-Cre+ PVH neurons and is likely quite heterogeneous with enriched expression of OTP, which is known to act in parallel with SIM1 in the specification of PVH neurons including AVP+, OXT+, and CRH+ and TRH+ neurons (, , , ). This group of neurons seems to include most “feeding regulatory” neurons described before. As shown in previous studies (, ), our scRNA-seq analysis revealed that AVP+ and OXT+ neurons are specifically compromised by Sim1 haploinsufficiency, whereas other neuron 2 subtypes (e.g., CRH+ and TRH+ neurons) are not as sensitive to the dosage of Sim1 (Fig. 1F). We detected a drastic reduction of Otp expression in the PVH neurons of neonatal Sim1Het mice (Fig. 1F and fig. S3A), although Otp expression was previously shown to be unaffected by KO of Sim1 (). Similarly, a significant reduction of Nr2f2, which encodes Chicken ovalbumin upstream promoter transcription factor II (COUP-TFII) that is coexpressed with POU3F2 in differentiating PVH neurons (), was also found in these PVH neurons of Sim1Het mice, despite the observation that only a slight reduction of Pou3f2 expression was detected. Thus, our data illustrate that Sim1 haploinsufficiency could lead to altered expression of Otp and Nr2f2, contributing to the specific PVH neuronal defects.

In this study, we have identified a poorly characterized population of PVH neurons (neuron 1), which constitutes about 20% of Sim1-Cre+ PVH neurons and appears to be a homogeneous group with high expression of TFs (Foxb1, Lhx1, and Pou2f2) and little or no expression of Otp, Nr2f2, and Pou3f2. These neurons also express a distinct set of neuropeptides including Cck, Cartpt, and Nts. These neuropeptide-expressing neurons have previously been detected in the PVH (–), but their functions are mostly linked to stress and inflammation. Nonetheless, CART+ PVH neurons could be activated by intraperitoneal injection of Cholecystokinin octapeptide (CCK-8S) in rats, suggesting that these neurons may play a role in the anorexigenic effects of peripheral CCK signals (). We found that the number of these PVH neurons is augmented by ~100% in Sim1Het mice. Further studies of these PVH neurons in feeding regulation are warranted.

In closing, we illustrated here abnormal hypothalamic activities of Irx3 and Irx5 as a central mechanism disrupting the development of PVH neurons leading to defective feeding regulation in Sim1 haploinsufficiency. However, there are many remaining questions. Additional studies will be needed to decipher how haploinsufficiency of Sim1 triggers the ectopic PVH expression of Irx3 and Irx5 and to determine whether ectopic expression of Irx3 and/or Irx5 in the PVH alone is sufficient to induce the neurodevelopmental defects of Sim1Het mice. By unveiling the abnormal hypothalamic function of IRX3 and IRX5, our work could also provide potential targets for obesity prevention and treatment.

MATERIALS AND METHODS

Animals

All animal experimental protocols approved by the Animal Care Committee of The Toronto Centre for Phenogenomics conformed to the standards of the Canadian Council on Animal Care. Sim1Het mice (), Irx3 mice (), Irx5 mice, Irx3−5 mice, and Irx3-floxed (Irx3) mice () were described previously. Sim1-Cre (stock no. 006451) and Ai14 (stock no. 007914) mice were obtained from the Jackson Laboratory. Mice were backcrossed for 10 generations on an outbred CD1 background. To generate Sim1-Cre;Irx3Fl/Fl mice, Irx3Fl/Fl mice were mated with Sim1-Cre mice. These mice were bred to Sim1Het mice to produce Sim1Het;Sim1-Cre;Irx3Fl/Fl mice. The colony was housed in a specific pathogen–free facility in ventilated cages with controlled environment settings (21° to 22°C, 30 to 60% humidity for standard housing), 12-hour light/12-hour dark cycles, and free access to water. Mice were randomly assigned to the experimental groups at the time of weaning.

Flow cytometry analysis and sorting

For Sim1-Cre;Ai14 tdTomato cell sorting, pups (P0 to P4) were dissected, and then hypothalami were rapidly extracted excluding optic chiasm and supra optic nucleus. Pooled hypothalamus samples from five to six mice were digested by papain solution (Worthington Biochemical) with deoxyribonuclease in Hibernate A medium for 30 min at 37°C with gentle agitation. Digestion was stopped by dilution with Hibernate A containing B27 (Thermo Fisher Scientific), GlutaMAX (Thermo Fisher Scientific), and 5% Trehalose (Sigma-Aldrich) and then gently triturated using a series of three Pasteur pipettes with tips fire-polished to incrementally smaller openings. Cell suspension was centrifuged at 300g for 5 min and resuspended in Hanks’ balanced salt solution with 0.5% bovine serum albumin (BSA) and 5% Trehalose. Sytox blue (Thermo Fisher Scientific; 1:1000) was added to the cell suspension, and cells were filtered a second time through a 70-μm cell strainer. Viable (Sytox blue–negative) tdTomato+ cells were subsequently sorted using a MoFlo Astrios (Beckman Coulter) cell sorter. Following cell sorting, cells were pelleted at 400g for 5 min, resuspended in phosphate-buffered saline (PBS) with 0.01% BSA, and counted with a hemocytometer. Total RNA from Sim1-Cre+ marked cells was extracted using the RNeasy Micro Kit (Qiagen).

Single-cell transcriptome sequencing

Following cell sorting, cells were pelleted at 400g for 5 min, resuspended in PBS with 0.01% BSA, and counted with a hemocytometer. Cells were adjusted to a final concentration of 140,000 cells/ml. Drop-seq was then carried out at the Princess Margaret Genomics Facility (Toronto, ON) as described previously (), except that flow rates of 3000 (cells and beads) and 13,000 μl/hour (oil) were used, and beads were resuspended at 167,000 per ml. Following Drop-seq droplet collection, complementary DNA amplification and sequencing library preparation were performed as described (), and the libraries were sequenced on a high-output Illumina NextSeq 500 flow cell at the Princess Margaret Genomics Centre (Toronto, ON).

scRNA-seq read alignment

Raw paired-end sequencing reads were processed using the Drop-seq tools (version 1.13) as described previously (). Reads were aligned to the mouse genome (mm10) using STAR v2.6.0b () with default settings. Following the cell selection method described in the Drop-seq Alignment Cookbook (version 1.2, written by James Nemesh from the McCarroll Lab, http://mccarrolllab.org/dropseq/), the cumulative fractions of reads for each successively smaller cell barcode were plotted to estimate the number of cell barcodes to be extracted from each experimental group. The inflection points of the plots were determined as the number of cell barcodes to be extracted and used in downstream analysis. Digital gene expression (DGE) matrices with 2807 and 2476 cell barcodes were selected for control and Sim1Het experimental groups, respectively. To run a single integrated analysis of all cells, DGE matrices were merged together for downstream analysis.

scRNA-seq data analysis

Cell barcodes with at least 200 genes detected were selected for clustering analysis using the Seurat R package (https://satijalab.org/seurat/) (). Briefly, following quality control by mitochondrial gene percentage and removal of cell doublets, gene expression measurements for each cell were divided by total gene expression, multiplied by 10,000, and natural log–transformed. Highly variable genes were identified with default settings of Seurat and used for principal components analysis, following data scaling to remove unwanted sources of variation by regressing on the number of molecules detected per cell and mitochondrial gene percentage. Significant principal components were used as input for UMAP plotting. Cluster markers were identified using the FindAllMarkers function in Seurat. Cluster identities were assigned on the basis of the top highly expressed genes and previously identified the PVH neuronal specific markers (, –, , , , , ). All statistical analyses of comparisons between experimental groups were performed using the chi-square test of independence with Bonferroni-adjusted P values. A probability value of P < 0.05 was used as the criterion for statistical significance.

Metabolic phenotyping experiments

Body weight was measured every week from 3 weeks of age. Body composition was analyzed using an EchoMRI device (Echo Medical Systems). Food intake of individually housed mice was determined as the difference in weight between Food-In and Food-Out, accounting for food spillage onto the cage floor on subsequent weeks, measured weekly from 3 weeks of age. Energy expenditure and RER (RER = VCO2/VO2) were evaluated at 8 to 10 weeks of age by indirect calorimetry using the Oxymax Comprehensive Lab Animal Monitoring System (Columbus Instruments) over periods of 24 hours. Briefly, energy expenditure was calculated by multiplying oxygen consumption (VO2) by the calorific value (calorific value = 3.815 + 1.232 × RER) and analyzed using analysis of covariance (ANCOVA) with lean mass adjustment or lean mass normalization. We simultaneously measured the locomotor activity of each mouse using the break counts of infrared beams. For glucose tolerance tests, mice were subjected to intraperitoneal injection of glucose (1 mg/g of body weight) after fasting overnight (14 to 16 hours). Blood glucose levels were measured at the indicated intervals using a glucometer (Contour NEXT, Bayer HealthCare).

Gene expression analyses by qPCR

Total RNA was extracted from tissue using the Qiagen RNeasy Mini or Micro Kit (Thermo Fisher Scientific), and complementary DNA was synthesized using Moloney-murine leukemia virus (M-MLV) reverse transcriptase (Thermo Fisher Scientific) with oligo(dT). Gene expression assay was conducted using SYBR Green methods on Viia7 (Applied Biosystems), and relative cycle threshold values were normalized by Actb (β-actin). The primers used for qPCR are shown in table S1.

In situ hybridization

Tissues were fixed overnight in 4% paraformaldehyde (PFA) at 4°C, processed, embedded in Optimal cutting temperature (OCT) compound, and sectioned at 20-μm thickness using cryostat (Leica) or embedded in paraffin and sectioned at 5-μm thickness using microtome (Leica). Section in situ hybridization against Irx3, Irx5, and tdTomato mRNA was carried out using published procedures with digoxigenin (DIG)–labeled RNA probes and anti–DIG-Peroxidase (POD) antibody (Roche; 1:2000) or fluorescein isothiocyanate–labeled RNA probes and anti–fluorescein isothiocyanate–Peroxidase (POD) antibody (Roche; 1:500) (, , ). The probes used for in situ hybridization are shown in table S1.

Histology

Tissues were fixed in 4% PFA and embedded in paraffin. Sections of 5 μm were stained with hematoxylin and eosin. Images were captured using a Nikon Eclipse microscope.

Immunostaining

Mice were deeply anesthetized with isoflurane and transcardially perfused with 1% PFA, and brains were harvested and fixed with 4% PFA. Fixed tissues were washed with PBS, embedded in OCT after cryoprotection with 30% sucrose, and then sectioned at 20-μm thickness using a cryostat (Leica) or 60 μm thickness using a vibratome (Leica). After antigen retrieval in 10 mM sodium citrate buffer (pH 6.0), brain slices were incubated for 1 hour at room temperature with 0.3% Triton X-100 in PBS (PBST) solution containing 5% donkey serum (Sigma-Aldrich) for blocking and permeabilization. After blocking, the samples were incubated overnight at 4°C with the following primary antibodies: anti-AVP (Millipore, AB1565, Rabbit; 1:5000), anti-NOS1 (Santa Cruz Biotechnology, A-11, Mouse; 1:200), anti-OXT (Millipore, AB911, Rabbit; 1:5000), anti-CART (Pheonix Pharmaceuticals, H-003-62, Rabbit; 1:2000), and anti–c-FOS (Santa Cruz Biotechnology, SC-52, Rabbit; 1:200). After several washes in PBST, tissues were incubated for 2 hours at room temperature with the following secondary antibodies: donkey anti-rabbit Alexa Fluor 488, 555, or 647 (Thermo Fisher Scientific; 1:2000) and donkey anti-mouse Alexa Fluor 488, 555, or 647 (Thermo Fisher Scientific; 1:2000). Slices were counterstained and mounted on the slide with mounting medium (Abcam), and then fluorescence signals were visualized using a Nikon A1R confocal microscope system and analyzed by ImageJ software (National Institutes of Health).

Neuronal activation by melanocortin agonist (MTII)

Six- to 7-week-old mice were fasted for 18 hours, injected with MTII (5 mg/kg; Canada Peptide), and then perfused 2 hours later with 1% PFA. Nuclear c-FOS immunoreactivity was then used as a marker for neuronal activation in PVH subsets.

Statistical analysis

Statistical analyses were performed using PRISM 6.0 software (GraphPad). Data are expressed as the means ± SEM values. Statistical significance of differences among groups was determined by Student’s t test or one-way analysis of variance (ANOVA) with post hoc analysis. When ANOVA indicated statistical significance, Tukey’s honestly significant difference test was then used to determine the significantly different means. A probability value of P < 0.05 was used as the criterion for statistical significance.