This study was approved by the local Animal Care and Use Committee of the Institute of Molecular Genetics CAS and the Institute of Anatomy, First Faculty of Medicine, Charles University. All experiments were performed with embryo littermates (females and males). Animals were housed in a controlled environment with 12-h light/dark cycles and free access to water and food.

The previously described experimental model of the conditional deletion of Hif1a (Hif1aCKO) genotype Isl1^tm1(cre)Sev/⁺;Hif1a^loxP/loxP was used (29). Briefly, floxed Hif1^atm3Rsjo with exon 2 of the Hif1a gene flanked by loxP sites (44) on a mixed C57BL/6J;C57BL/6N genetic background were obtained from Jackson Laboratories (Bar Harbor, ME, USA; #Strain 007561). Isl1^Cre/+ mice were on the FVB background. Hif1a^loxP/+ or Hif1a^loxP/loxP mice without the Isl1-Cre allele individuals were used as the control. For the breeding scheme, female mice Hif1a^loxP/loxP were crossed with Hif1a^loxP/+; Isl1^Cre/+ males, in which the Isl1-Cre knock-in allele was inherited paternally to minimize the potential influence of maternal genotype on the developing embryos.

To visualize sympathetic neurons, we used tdTomatoAi14 reporter mice with Rosa-CAG-LSL-tdTomato allele (Ai14, B6.Cg-Gt(ROSA)26Sor^{tm14(CAG-tdTomato)Hze} , Stock No. 7914 Jackson Laboratories) to generate the reporter Hif1aCKO-Ai14 (genotype: Hif1a^loxP/loxP;Isl1^Cre;tdTomatoAi14) and control-Ai14 mice (genotype: Hif1a^loxP/+;Isl1^Cre;TdtomatoAi14). To examine the formation of the secondary sympathetic chain, we used peripherin-enhanced green fluorescent protein (Prph-eGFP) genomic reporter transgenic mice (45, 46) and crossed them with the Hif1a mutant line.

To visualize the pattern of the developing coronary arteries, Hif1a^loxP/loxP mice were crossed to Cx40:eGFP strain with eGFP signal in the coronary arteries, the atria, the atrioventricular node, and the His–Purkinje system (47). Double homozygote females (Cx40:^eGFP/+ ;Hif1a^loxP/loxP ) with background CD1/129SvJxSwiss were then crossed with Hif1a^loxP/+;Isl1^Cre males.

The noon of the day on which the vaginal plug was found was designated E0.5. Animals were euthanized by cervical dislocation in our study. Embryos were collected for analysis at different ages ( Figure 1A ). All comparisons were made between animals with the same genetic background. Phenotyping and data analysis were performed blind to the genotype of the mice. Genotyping was performed by PCR on tail DNA ( Supplementary Table 1 ).

Diabetes was induced in 6-week-old females (Hif1a^loxP/loxP ) by two intraperitoneal injections of 100 mg/kg body weight of streptozotocin (STZ, S0130, Sigma-Aldrich, Dorset, UK) within a 1-week interval, as described (37, 48). The level of blood glucose was checked in a drop of blood from the tail vessel using a glucometer (CONTOUR plus ONE, Ascensia Diabetes Care, Basel, Switzerland). Mice with a level of glucose maintained above 13.9 mmol/l in blood were classified as diabetic. Diabetic females were mated with males, and the next morning, the vaginal plug was checked. Maternal blood glucose levels at the time of embryo collection are shown in Supplementary Figure 1 . Embryos collected from diabetic pregnancies were labeled as "diabetic" or "DIA", and from non-diabetic pregnancies were labeled as "non-diabetic" or "non-DIA".

Whole-mount immunohistochemical staining of embryonic hearts (E16.5 and E18.5) was performed on cleared tissue, as described (49). Embryos were perfused with 0.05% heparin in phosphate-buffered saline (PBS) and fixed in 4% paraformaldehyde (PFA) for 90 minutes. The heart was cleared using CUBIC reagent at 37°C with gentle shaking for 1 week (50). Used primary and secondary antibodies are described in Supplementary Tables 2 and 3 . The nuclei were stained with Hoechst. The fluorescent signals were detected using LSM 880 NLO (AxioObserver Z1, Carl Zeiss, Oberkochen, Germany) and AxioZoomV16 (Carl Zeiss, Germany).

Positive areas (TH and TUJ1) were quantified using the Threshold tool in NIH ImageJ software and expressed as a percentage of total areas (heart and adrenal gland). The total number of positive cells (NeuN) in the whole stellate ganglion (STG) was quantified using a cell counter in NIH ImageJ software. Microvasculature density was assessed using immunohistochemically stained 8-µm paraffin sections labeled with anti-PECAM-1 (an endothelial marker) and wheat germ agglutinin (WGA). Sections were imaged on an Olympus confocal microscope, and positive signals of PECAM-1 and WGA were combined and thresholded in NIH ImageJ. Then, the positive area was measured in the region of interest (compact layer) in the right ventricle. Subsequently, the measured microvasculature-positive area was normalized to the background signal (all tissue autofluorescence) and expressed as a percentage, thus leading to the density of microvasculature in the right ventricular wall.

To visualize the pattern of the developing coronary arteries together with TH⁺ innervation, the hearts were dissected and immunolabeled with anti-TH and Cy5-conjugated secondary antibodies. The atria were then carefully cut off to expose the origin of the coronary arteries at the base of the heart, cleared in CUBIC for 48 h (51), and visualized with 4× and 10× dry objectives on an Olympus FluoView 1000 confocal system. For quantification of TH⁺ innervation, maximum intensity projections (MIPs) of confocal series taken with a 4× objective with 25-μm z-step were used. The image from the far-red (Cy5) channel containing the signal from the anti-TH antibody was oriented with the base on the top and the apex at the bottom (Valentine projection). Two semicircular segmented lines at 25% and (basal) and 75% (apical) apex–base distance were drawn in ImageJ, and line profiles were then generated ( Supplementary Figure 2 ). The peaks in fluorescence intensity, corresponding to individual nerve bundles, were then easily counted in an unbiased manner.

Embryonic hearts were microdissected from non-diabetic and diabetic control and Hif1aCKO embryos (E16.5 and E18.5). An advanced CUBIC protocol (49) with some modifications (50) was used for tissue clearing to enable efficient imaging by light-sheet microscopy. Whole-mount immunohistochemical staining of embryonic hearts was performed, and samples were stored before imaging in Cubic 2 at room temperature. The secondary sympathetic chain was microdissected from non-diabetic and diabetic control-Ai14 and Hif1aCKO-Ai14 embryos (E14.5) and whole-mount stained with NeuN. Zeiss Lightsheet Z.1 microscope with illumination objective Lightsheet Z.1 5×/0.1 and detection objective Dry objective Lightsheet Z.1 5×/0.16 was used for imaging at the Light Microscopy Core Facility of the Institute of Molecular Genetics of the Czech Academy of Sciences. IMARIS software v8.1.1 (Bitplane AG, CA, San Francisco, USA) was used for image processing.

Sympathetic ganglia were dissected from E14.5 embryos Hif1aCKO-Ai14;Prph-eGFP (n = 3) and Control-Ai14;Prph-eGFP (n = 2). Sympathetic chains were homogenized, and neuronal cells were dissociated using 0.1% collagenase (C9263, Sigma-Aldrich, UK) and 0.05% trypsin in Dulbecco’s PBS for 7 minutes (T4799, Sigma-Aldrich, UK). Enzymatic activity was stopped by adding fluorescence-activated cell sorting (FACS) buffer [2% fetal bovine serum (FBS) in Dulbecco’s PBS and 10 mM EGTA]. FACS of eGFP and tdTomato-positive cells was performed at the Imaging Methods Core Facility at BIOCEV on a BD FACS Aria Fusion flow cytometer operated using BD FACSDiva™ Software. A total of 100 eGFP⁺ and tdTomato⁺ cells per biological sample were collected into individual wells of a 96-well plate containing 5 µL of lysis buffer of NEBNext single-cell low input RNA library prep kit for Illumina (#E6420, New England Biolabs, Ipswich, MA, USA). Plates were frozen immediately on dry ice and stored at −80°C. The total time from euthanasia to cell collection was ∼3 h.

The RNA library preparation, RNA sequencing (RNA-seq), and data processing were performed as described previously (27). Briefly, the NEBNext single-cell low-input RNA library prep kit for Illumina (#E6420, New England Biolabs) was used for library generation at the Gene Core Facility (Institute of Biotechnology CAS, Czechia), and the libraries were sequenced on an Illumina NextSeq 500 next-generation sequencer with NextSeq 500/550 High Output kit 75 cycles (Illumina #200024906) at the Genomics and Bioinformatics Core Facility (Institute of Molecular Genetics CAS, Czechia). RNA-seq reads in FASTQ files were mapped to the mouse genome GRCm38 primary assembly release M8 using STAR [version 2.7.0c (52)]. Using cutadapt v1.18 (53), the number of reads (minimum, 32 million; maximum, 73 million) was trimmed by Illumina sequencing adaptor and bases with reading quality lower than 20; subsequently, reads shorter than 20 bp were filtered out. TrimmomaticPE version 0.36 (54). Ribosomal RNA and reads mapping to UniVec database were filtered out using bowtie v1.2.2. with parameters -S -n 1 and SortMeRNA (55). A count table was generated using the Rsubread v2.0.1 package with default parameters without counting multi-mapping reads. The raw RNA-seq data were deposited at NIH GEO under accession number GSE250606 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE250606).

DESeq2 [v1.26.0 (56)] default parameters were used to normalize data and compare the different groups. Differentially expressed genes between Hif1aCKO and control sympathetic neurons were identified based on an adjusted p-value p_adj < 0.05, log2 fold change (log2FC >0.3, <−0.3), and a base mean ≥50. The functional annotation of the differentially expressed genes was performed using g: Profiler (Raudvere et al., 2019). Complete query details are available in Query info Tables in Dataset S1. The resulting GEM and combined GMT files using term size <1,800 were loaded into Cytoscape (57) plugin “EnrichmentMap” (58) using 0.01 false discovery rate (FDR) q-value cutoff to generate a network. The edge cutoff was set to 0.35, and nodes were set to 0.007 Q value.

RNA was isolated from the microdissected sympathetic chains of individual embryos from diabetic and non-diabetic litters using TRIzol (Invitrogen, Carlsbad, CA, USA). The concentration and purity were quantified using NanoDrop (ND-2000 Spectrophotometers, Thermo Fisher Scientific, Waltham, MA, USA). cDNA samples were prepared using Maxima H Minus First Strand cDNA Synthesis Kit with dsDNA (#K1682, Thermo Scientific, USA) from 300 ng isolated RNA/sample. Quantitative real-time PCR (qRT-PCR) was performed using 10× diluted cDNA samples. cDNA at a volume of 4 µL was added to 5 µL of SybrGreen (GrandMaster Mix, TATAA Biocenter, Gothenburg, Sweden) with 0.2 µM reverse and forward primers. Primers were designed using Primer3 software, and sequences are shown in Supplementary Table S2 . Validation of RNA-seq targets was performed using Bio-Rad C1000 Thermal Cycler (CFX384 Real-Time System, Bio-Rad Laboratories, Hercules, CA, USA), and activation was performed using AmpliTaq at 95°C for 10 minutes, followed by 40 cycles at 95°C for 15 s for denaturation and 60°C for 60 s for extension. The relative expression levels of mRNA of target genes were normalized to the reference gene Hprt1. All reactions were conducted in duplicates, and the data were calculated using the ΔΔCp method, as previously described (37, 59). Primer sequences are presented in Supplementary Table 4 .

Statistical analyses were performed using two-way ANOVA (GraphPad Prism 10), testing differences among experimental groups based on the genotype and experimental condition (diabetic or non-diabetic pregnancy) followed by multiple Fisher’s comparisons; results are expressed as mean ± SD or mean ± SEM, with significance level p < 0.05. A chi-square (χ²) test was used to compare the distribution of genotypes between the non-diabetic and diabetic groups (GraphPad Prism 10). Sample sizes and individual statistical results for all analyses are provided in the figure legends and tables.

We analyzed the combined impact of the Hif1a-deficient cardiac sympathetic system and the adverse maternal diabetes environment on heart development (schematics of experimental study design in Figure 1A ). The observed distribution of embryo genotypes collected at E14.5 was influenced by maternal diabetes (**p = 0.0012, χ² test; Figure 1B ). We did not observe any developmental delay or structural abnormalities among embryos from diabetic and non-diabetic pregnancies. However, we identified a significant decrease in the average crown–rump length of both diabetic control and Hif1aCKO embryos compared to non-diabetic controls at E14.5 ( Figures 1C, D ). Furthermore, the length of diabetic Hif1aCKO embryos was significantly smaller than that of diabetic controls or non-diabetic Hif1aCKO, indicating a synergistic effect of Hif1a deletion and the diabetic environment.

To assess the extent of cardiac innervation, we employed double immunolabeling of TH, a marker of sympathetic neurons and sympathetic innervation, and class III β-tubulin (TUJ1), a neuronal marker expressed in all cardiac fibers, representing sympathetic, parasympathetic, and sensory innervation. While parasympathetic cardiac innervation precedes sympathetic innervation, sympathetic fibers move alongside established vagal nerve tracts to innervate the heart (60). In the developing mouse heart, autonomic innervation precedes sensory innervation, with sensory axons becoming detectable at E18.5 (6, 61). In line with our previous study (29), the conditional deletion of Hif1a in sympathoadrenal progenitor lineage led to a profound deficit in cardiac sympathetic innervation ( Figures 2A–D ; Supplementary Video S1 - S4 ). At E16.5, our immunohistochemical staining revealed that the majority of TH⁺ axons in the posterior part of the ventricles of diabetic Hif1aCKO were lost, with no cardiac fibers observed in the apex ( Figures 2B, D ). Maternal diabetes and Hif1a mutation significantly reduced TH⁺ and TUJ1⁺ cardiac innervation when compared to hearts from non-diabetic pregnancies and control hearts, respectively ( Figures 2D, E ). Notably, a negative synergistic effect of Hif1a deletion and the diabetic environment on cardiac innervation was observed when comparing Hif1aCKO and control hearts from diabetic pregnancies.

Next, we evaluated the sympathetic innervation of the posterior and anterior parts of the ventricular wall in detail. Sympathetic innervation, as marked by anti-TH, was notably reduced in the Hif1aCKO ventricles, with a loss of both proximal and distal branches compared to the control hearts at E17.5, mirroring the observation at E16.5 ( Figure 3A ). The difference at the base of the heart, where the thick bundles dominated, was less affected between genotypes; however, a clear decrease in the number of (usually thinner) bundles was evident at the apex, attesting to a deficient innervation. The maternal diabetes exposure resulted in a significant attenuation of sympathetic innervation in the control hearts ( Figures 3A, B ). However, the synergistic detrimental effects of Hif1a deletion and diabetic pregnancy were most pronounced in the diabetic Hif1aCKO hearts. Only a few main branches innervated the basal area of the myocardial wall in the anterior view, and all distal ventricular branches were missing. While the posterior part of the heart contained a higher number of branches with a more complex plexus, the combination of Hif1a deletion and maternal diabetes led to a nearly complete loss of innervation, with only a few remaining nerves ( Figures 3A, B ).

At E18.5, the most pronounced deficit in cardiac sympathetic innervation was detected in Hif1aCKO embryos exposed to maternal diabetes, indicating a long-lasting synergistic effect of Hif1a deletion and the diabetic environment ( Figures 4A–D ). Conversely, TUJ1⁺ axons were equally reduced in both non-diabetic and diabetic Hif1aCKO hearts, highlighting the impact of Hif1a mutation rather than maternal diabetes ( Figure 4E ). Similarly, exposure to maternal diabetes did not significantly affect TUJ1 innervation in the control heart compared to control hearts from non-diabetic embryos. Both sympathetic and TUJ1⁺ fibers were significantly reduced in non-diabetic Hif1aCKO hearts when compared to non-diabetic control hearts ( Figures 4D, E ), indicating abnormalities in the formation of cardiac innervation of Hif1aCKO. Consequently, these changes could have significant implications for heart function and postnatal survival of Hif1aCKO mutants.

Given the implication of the sympathetic nervous system in heart size regulation and cardiomyocyte proliferation (19, 20), we conducted an analysis of the wall thickness of both the left ventricle (LV) and right ventricle (RV) and the thickness of the interventricular septum, using the sections of the E17.5 heart ( Figure 5A ). First, the length of the heart (from apex to base) and width across the widest part of both ventricles were measured. Maternal diabetes and Hif1a mutation significantly reduced both measured parameters, although the ratio of length and width was not altered, indicating that the ventricular proportions were similar among groups ( Figures 5A, B ). The compact myocardium of the LV and RV of the Hif1aCKO heart was significantly thinner compared to those of their control littermates from non-diabetic pregnancies ( Figures 5A, C ). Additionally, the control embryos from the diabetic pregnancy exhibited thinner LV and RV walls when compared to control non-diabetic embryos. Interestingly, the thickness of the interventricular septum increased in both diabetic and non-diabetic Hif1aCKO and diabetic control hearts compared to the control heart of embryos from non-diabetic pregnancies ( Figures 5A, C ). However, we did not detect an additive effect of the combination of Hif1a mutation and the diabetic environment on these parameters when compared to diabetic control embryos.

Next, we investigated the coronary vascular architecture, as the development of arterial and sympathetic nerve networks is coordinated in terms of both spatial distribution and molecular signaling. Furthermore, our previous study revealed a higher neonatal mortality rate among Hif1aCKO mice, which could be partially linked to major coronary artery anomalies (29). Building upon these findings, our current study expands our analyses of the formation of coronary artery architecture, using the Cx40:eGFP knock-in model with eGFP signal in the coronary arteries (47). The normal pattern of the coronary arteries together with examples of abnormal findings recorded from the mutant and diabetic hearts is summarized in Figure 6 . Abnormalities included multiple smaller branches instead of a single large one, arterial “windows”, and anomalous connections to the aorta (separate orifices of the circumflex and left anterior descending branches of the left coronary artery and separate orifice of the septal branch of the right coronary artery) instead of single opening of the left and right coronary arteries. These abnormalities alone cannot account for the approximately 40% perinatal mortality observed in the Hif1aCKO mice (29). Thus, the prenatal evaluations did not uncover any new, more severe malformations in the coronary artery architecture in the Hif1aCKO embryos, in comparison to the analyses of the adult Hif1aCKO heart reported previously (29).

We next visualized the microvasculature within the heart. We used a combination of anti-PECAM-1 antibody (a pan-endothelial marker) and WGA labeling. Subepicardial veins were noticeably dilated in both non-diabetic and diabetic Hif1aCKO in the RV (white arrows in Figures 7C, D ) when compared to the control hearts from diabetic and non-diabetic embryos ( Figures 7A, B ). Interestingly, coronary arteries in the compact myocardium of the RV appeared dilated in embryos from diabetic pregnancies (yellow arrows in Figures 7B, D ) but not in non-diabetic Hif1aCKO hearts. Although the relative vessel density, indicative of myocardial perfusion and oxygenation in the RV, did not show significant differences among the groups ( Supplementary Figure 3 ), the thinner ventricular wall represents a reduction in the absolute amount of microvasculature compared to the controls. The microvasculature of the LV seemed to be less affected, although subepicardial veins were also dilated in diabetic and non-diabetic Hif1aCKO and in diabetes-exposed control embryos compared to the non-diabetic control group ( Figures 7E–H ).

Sympathetic neurons and neuroendocrine chromaffin cells in the adrenal medulla originate from a common catecholaminergic sympathoadrenal progenitor (62). Therefore, we proceeded to examine the development of chromaffin cells in the adrenal medulla. Consistent with our earlier findings that HIF-1α deficiency adversely impacted the development of chromaffin cells in the adrenal medulla (29), we observed a significant reduction in TH expression in the adrenal medulla Hif1aCKO embryos as early as E14.5 ( Figures 8A, B ). However, the detrimental effect of maternal diabetes on TH-expressing chromaffin cells became evident at a later stage, specifically at E18.5. Maternal diabetes exposure led to a significant reduction in the size of the adrenal medulla of control embryos, but diabetic Hif1aCKO embryos exhibited the most substantial reduction when compared to the other groups.

Considering the compromised cardiac sympathetic innervation resulting from HIF-1α deficiency and exposure to a maternal diabetic environment, we assessed the development of postganglionic neurons. The stellate ganglia and thoracic sympathetic chain were evaluated using light-sheet fluorescence microscopy and tdTomato and Prph-eGFP reporter expression at E14.5 ( Figures 9A, B ). The size of the sympathetic chain was reduced in response to maternal diabetes and Hif1a deletion when compared to non-diabetic control sympathetic chains ( Figure 9D ). Similarly, the number of neurons expressing NeuN, a marker of mature neurons, exhibited a decrease in both maternal diabetes and Hif1a mutation conditions, with no observed additive combinatorial effect of Hif1a mutation within the diabetic environment ( Figures 9C, E ).

To analyze the molecular changes resulting from the elimination of Hif1a in developing sympathetic neurons, we performed a bulk-cell RNA sequencing (experimental design in Figure 10A ). Each biological replicate for the bulk RNA-seq analysis contained 100 Prph-eGFP⁺ and tdTomato⁺ FACS-sorted single cells from the E14.5 stellate ganglia and thoracic sympathetic chain.

Differential expression analysis identified 36 downregulated and 32 upregulated transcripts of protein-coding genes in Hif1aCKO neurons ( Supplementary Data 1 ). Functional enrichment analysis of the set of upregulated genes demonstrated a significant enrichment of Gene Ontology (GO) categories associated with proliferation, cell cycle, and mitosis, likely reflecting the main compensatory mechanisms for aberrant neuronal development ( Figures 10B, C ; Supplementary Data 1 ). For example, upregulated genes encoding proteins are involved in cell cycle and mitosis regulation including Cdc20, the coactivator of mitotic progression, which is also required for dendrite development (63); the transcription factor E2f1, which is important for cell cycle progression and apoptosis (64); cyclin B1 (Ccnb1) (65); Kif23, which is important for cell division and implicated in neuronal migration (66); Cenpe, Cenpf, and Mki67, which are important for cell cycle and mitosis (67); and Mcc, which is a regulator of cell cycle and Wnt/β-catenin signaling pathway (68). Additionally, we detected an elevated expression of glutamic acid decarboxylase 2 (Gad2), the enzyme that catalyzes the formation of γ-aminobutyric acid (GABA) from glutamic acid. GABAergic signaling, the main inhibitory neurotransmitter system, is important in sympathetic and cardiovascular regulation (69). Slc38a11, a member of the SLC38 family of amino acid transporters, was upregulated, although its function and substrate specificity are unknown (70). Moreover, elevated levels of Gata2, a key transcription factor expressed in developing sympathetic neurons (71), were found in Hif1aCKO neurons.

The GO analysis of downregulated genes revealed a notable enrichment of biological processes linked to nervous system development, neuron death, secretion, and synapse function, indicating significant alternations in neuronal development, survival, and function. The downregulated set included genes such as HIF-1, regulated and highly expressed in neurons glucose transporter (Slc2a3) (72), the neuronal calcium sensor neurocalcin delta (Ncald) (73), a synaptic organizer and adhesion molecule neurexin 1 (Nrxn1) (74), and neurotrophic receptor tyrosine kinase 1, Ntrk1 (also known as TrkA), a receptor for nerve growth factor with a key role in the regulation of proliferation, differentiation, and survival of sympathetic neurons (75, 76).

Using our RNA-seq data, we selected specific genes for qRT-PCR analysis to examine their expression in the sympathetic chain ganglia of control and Hif1aCKO embryos from non-diabetic and diabetic pregnancies at E14.5 (experimental design in Figure 10A ). Consistent with the RNA-seq results, we found significantly increased expression levels of several selected genes, including Cdc20, Gad2, Gata2, Kif23, Mki67, and Scl38a11, in Hif1aCKO compared to the control group ( Figure 10D ). Additionally, some of these genes (Gad2, Kif23, and Mki67) were also upregulated in the sympathetic chain ganglia of control embryos exposed to maternal diabetes. While there was no additive effect of maternal diabetes in conjunction with Hif1a deletion on their expression, these results suggest possible compensatory mechanisms associated with proliferation, cell cycle, and mitosis in response to the adverse diabetic environment or Hif1a deficiency in developing sympathetic neurons. Interestingly, genes associated with synapse function, such as Nrxn1 and Ncald, exhibited reduced expression in both control and Hif1aCKO embryos from diabetic pregnancies. In contrast, Plk1 encoding Polo-like kinase 1, a regulator of mitosis (77), was upregulated by maternal diabetes but not affected by Hif1a deletion. Overall, these results align with our analyses of cellular changes in sympathetic chains, with no observed additive combinatorial effect of Hif1a mutation in the diabetic environment.