Depletion of the isocitrate dehydrogenase 3 subunit IDHA-1 licenses germ cell-to-neuron direct reprogramming

Abstract

Direct reprogramming (DR) converts one differentiated cell identity into another and is a promising strategy for regenerative approaches when cells are lost because of injury or disease. Transcription factors (TFs) that specify a given cell type can drive DR, but inhibitory mechanisms restrict this conversion in most cell types. To identify such barriers in vivo, we use Caenorhabditis elegans and the zinc finger TF CHE-1, which is required to specify the glutamatergic taste neuron fate of ASE neurons. Ectopic CHE-1 expression can induce DR of different cell types to ASE neurons upon RNAi-mediated depletion of barrier genes.

Here we characterize the α subunit of the mitochondrial isocitrate dehydrogenase 3 complex (IDH3), IDHA-1, as a barrier to DR of germ cells into neurons. RNAi against idha-1 produced consistent ASE fate reporter expression, accompanied by morphological changes to neuron-like structures in the germline upon ectopic CHE-1 expression. Using different neuronal gene expression reporters, single-molecule FISH, and antibody staining, we confirm that this germ cell conversion (GeCo) produces neuron-like cells. We found that loss of IDH3 activity causes metabolic perturbations that result in a variety of direct and indirect effects on DR of germ cells. One arm implicates the hypoxia-inducible factor HIF-1, which promotes GeCo. Its loss causes a significant decrease, while the vhl-1 mutant background, in which HIF-1 protein is stabilized, leads to enhanced GeCo. Another arm implicates epigenetic changes leading to a detectable loss of the repressive histone marks H3K27me3 and H3K9me3 upon idha-1 depletion. Furthermore, stable-isotope-resolved metabolomics shows strong citrate accumulation without a noteworthy drop in α-Ketoglutarate (αKG) levels, indicating that compensatory pathways maintain αKG levels. Genetics and metabolomics analyses confirmed that Glutamate anaplerosis contributes to compensating for the loss of IDHA-1. Additionally, we found that depletion of glucose transporter FDGT-2 nearly doubles GeCo efficiency. Overall, our findings define the TCA cycle as an in vivo safeguard of germ cell identity and show that multiple metabolic inputs converge to keep germ cells refractory to TF-induced DR to neuronal cells.

INTRODUCTION

Differentiated cell identity is maintained by mechanisms that prevent activation of alternative gene expression programs. Yet, transcription factors (TFs) can revert differentiated cells to pluripotency by reprogramming, or convert them directly into another differentiated type (Davis, Weintraub et al. 1987, Schneuwly, Klemenz et al. 1987, Takahashi and Yamanaka 2006, Vierbuchen, Ostermeier et al. 2010). Cellular reprogramming is typically inefficient or completely blocked, indicating that barriers safeguard resident identity. Hence, to induce cell fate conversion, a TF must overcome the cell’s molecular processes that restrict reprogramming (Soufi and Zaret 2013, Becker, Nicetto et al. 2016, Brumbaugh, Di Stefano et al. 2019). Identifying these barriers helps clarify how cellular identity is maintained and can make direct reprogramming (DR) more efficient for regenerative applications. Previously characterized barriers often relate directly to chromatin regulation and include histone chaperones, chromatin remodellers, histone-modifying enzymes, and complexes that maintain chromatin states. They have repeatedly emerged from unbiased screens for barriers using TF-induced fate conversion in both invertebrate and vertebrate systems (Tursun, Patel et al. 2011, Onder, Kara et al. 2012, Cheloufi, Elling et al. 2015, Kolundzic, Ofenbauer et al. 2018, Hajduskova, Baytek et al. 2019). A less explored class of molecular processes that affects cell identity is cellular metabolism. Reprogramming somatic cells into induced pluripotent stem cells (iPSCs) shifts metabolism from oxidative phosphorylation to glycolysis, and manipulating this shift alters reprogramming efficiency (Folmes, Nelson et al. 2011, Zhang, Nuebel et al. 2012, Zhang, Wang et al. 2015). Metabolism is mechanistically linked to chromatin state by delivering substrates or cofactors of chromatin-modifying enzymes such as acetyl-CoA for histone acetyltransferases, S-adenosylmethionine for methyltransferases, and α-Ketoglutarate (αKG) for dioxygenases that include histone demethylases and protein hydroxylases. In particular, αKG which is produced in the mitochondrial matrix by isocitrate dehydrogenase, were shown to control the pluripotent state of mouse embryonic stem cells (Carey, Finley et al. 2015, Tischler, Gruhn et al. 2019) accelerate differentiation of primed human pluripotent stem cells (TeSlaa, Chaikovsky et al. 2016), and prevent transformation of fibroblasts into cancer-associated fibroblasts (Zhang, Wang et al. 2015). Also, succinate and fumarate act as competitive inhibitors of αKG-dependent dioxygenases thereby blocking cellular transformation to glioma and leukemia (Figueroa, Abdel-Wahab et al. 2010, Xu, Yang et al. 2011, Lu, Ward et al. 2012, Xiao, Yang et al. 2012). Whether perturbations at the αKG node can create permissiveness for direct reprogramming in vivo remains unclear because it requires a system in which fate conversion can be assessed quantitatively in an intact animal.

The nematode C. elegans serves as such a model system, providing a transparent body for fluorescent fate reporters with efficient RNAi and other molecular techniques. We previously performed a whole-genome RNAi screen for candidate barrier genes using the ectopic expression of the Zn-finger TF CHE-1 (Kolundzic, Ofenbauer et al. 2018). CHE-1 is the terminal selector of the glutamatergic ASE taste neuron (Chang, Johnston et al. 2003, Etchberger, Lorch et al. 2007), but is alone insufficient to convert cells when overexpressed in other tissues. However, upon depletion of the histone chaperone LIN-53 (RBBP4/7) or FACT, germ cells are converted into ASE-like neurons that express the ASE marker gcy-5 and adopt neuron-like morphology (Tursun, Patel et al. 2011, Kolundzic, Ofenbauer et al. 2018). Removal of Polycomb repressive complex 2 components phenocopies this effect (Patel, Tursun et al. 2012), and subsequent work identified the chromodomain protein MRG-1 (Hajduskova, Baytek et al. 2019) as further germ cell identity safeguards. Importantly, the whole-genome RNAi screen recovered not only chromatin factors but also genes with no obvious connection to chromatin and gene regulation. Among them was idha-1, encoding the α subunit of the mitochondrial NAD-dependent isocitrate dehydrogenase 3 complex (IDH3) (Lin, Chang et al. 2022). RNAi against idha-1 allows overexpressed CHE-1 to induce the ASE fate reporter gcy-5::gfp in the germline, raising the possibility that the TCA cycle contributes to germ cell identity maintenance in vivo. Here, we show that depletion of the catalytic α or the regulatory γ subunit of IDH3, but not other isocitrate dehydrogenase genes, renders germ cells permissive for CHE-1-induced conversion into neuron-like cells.

Perturbing a central metabolic enzyme such as an isocitrate dehydrogenase can trigger multiple changes beyond altered TCA intermediates and amino acid pools, including changes in oxygen sensing, nuclear histone methylation, and many other cellular processes. Hence, the causal chain from loss of IDH3 activity to the permissive state is not straightforward to establish. Our data show that GeCo upon idha-1 depletion is promoted by HIF-1 and specific JmjC-domain proteins, and accompanied by loss of repressive histone methylation in the germline. Surprisingly, knockdown of idha-1 is not accompanied by a measurable decrease in bulk αKG but by the modulation of glutamine/glutamate anaplerosis and glucose transport. We therefore present a model in which several partially independent metabolic inputs converge on germline chromatin to allow TF-induced direct reprogramming.

RESULTS

Depletion of the IDH3 α or γ subunit permits CHE-1-induced germ cell conversion

To identify reprogramming barriers in vivo, we previously established animals carrying a heat-shock-inducible broad overexpression transgene of che-1 together with the ASE fate reporter gcy-5^prom::GFP (Tursun, Patel et al. 2011). CHE-1 overexpression alone is very limited in inducing gcy-5^prom::GFP, unless depletion of a barrier gene lifts this restriction (Tursun, Patel et al. 2011, Kolundzic, Ofenbauer et al. 2018) (Figure 1A). The gene idha-1, encoding a mitochondrial isocitrate dehydrogenase, was identified in a previous genome-wide RNAi screen (Kolundzic, Ofenbauer et al. 2018) but remained uncharacterized. Combining idha-1 RNAi with CHE-1 overexpression (CHE-1oe) induced gcy-5^prom::GFP expression in the germline, accompanied by morphological changes in which germ cells acquired neuron-like structures resembling axo-dendritic projections (Figure 1B).

Figure 1. Depletion of mitochondrial IDH3 subunits permits CHE-1-induced ASE fate reporter expression in the C. elegans germline.

(A) Schematic of the transcription factor-induced direct reprogramming system. Animals express the ASE fate reporter gcy-5^prom::GFP in the ASE neuron only. Broad overexpression of the ASE neuron fate-inducing transcription factor CHE-1 (CHE-1^OE) is insufficient to induce the ASE fate reporter in additional cells, indicating cell-fate safeguarding mechanisms that restrict ectopic CHE-1 activity. RNAi-mediated depletion of idha-1 decreases the barrier and permits ectopic gcy-5^prom::GFP expression in germ cells. (B) Representative images of animals carrying gcy-5^prom::GFP and subjected to broad CHE-1 overexpression following control or idha-1 RNAi. Ectopic GFP expression is absent from the germline in control RNAi-treated animals but is induced in germ cells following idha-1 depletion. These gcy-5^prom::GFP-positive germ cells also acquire a neuron-like morphology, including prominent neurite-like projections that are clearly distinct from the typical morphology of germ cells. Boxed regions are shown at higher magnification; arrowheads indicate ectopic gcy-5^prom::GFP positive germ cells and associated projections. (C) Schematic representation of the mitochondrial tricarboxylic acid (TCA) cycle. NAD+-dependent isocitrate dehydrogenase 3 (IDH3) catalyzes the conversion of isocitrate to α-ketoglutarate (α-KG). (D) Schematic of the heterotetrameric mammalian IDH3 complex, comprising two α-subunits, one β-subunit and one γ-subunit, with the corresponding C. elegans homologs IDHA-1, IDHB-1 and IDHG-1 indicated. (E) Quantification of animals displaying ectopic gcy-5^prom::GFP expression in the germline following RNAi-mediated depletion of the indicated IDH3 subunits and related isocitrate dehydrogenases in combination with CHE-1^OE. Depletion of idha-1 or idhg-1, encoding the α and γ subunits of mitochondrial NAD⁺-dependent IDH3, respectively, significantly increased ectopic germline GFP expression relative to control RNAi. In contrast, depletion of the isocitrate dehydrogenases idh-1 and idh-2 did not permit ectopic germline GFP expression. IDH-1 and IDH-2 are NADP+-dependent isocitrate dehydrogenases, distinct from the NAD+-dependent IDH3 complex that catalyses the canonical oxidative TCA cycle reaction.

Data are presented as mean ± SEM from three independent biological experiments. Statistical significance was determined by ordinary one way ANOVA followed by Dunnett’s multiple comparisons test, with each RNAi condition compared with control RNAi. ****P < 0.0001; ns, not significant. Numbers below the bars indicate the total number of animals scored.

IDHA-1 protein is part of the NAD⁺-dependent isocitrate dehydrogenase complex IDH3 of the mitochondrial matrix, which catalyzes the oxidative decarboxylation of isocitrate to αKG, in the TCA cycle (Gabriel, Zervos et al. 1986) (Figure 1C). In humans, the enzyme assembles as a heterotetramer of catalytic α subunits with regulatory β and γ subunits (Chen and Ding 2023); C. elegans genes encode one α subunit (IDHA-1), one β subunit (IDHB-1), and two γ subunits (IDHG-1 and IDHG-2) (Figure 1D). We tested RNAi against the corresponding genes and those encoding other isocitrate dehydrogenase genes, including the NADP-dependent enzymes IDH-1 and IDH-2. Only depletion of idha-1 (~30% of animals; n = 400) and idhg-1 (~16%; n = 600) allowed gcy-5^prom::GFP induction by CHE-1oe (Figure 1E). The barrier effect is therefore specific to the α and one of the two γ subunits of the mitochondrial NAD-dependent complex.

To assess which tissues IDHA-1 and IDHG-1 are expressed, we generated CRISPR knock-in alleles tagging IDHA-1 with 3xHA and IDHG-1 with 3xFLAG for immunostaining. Both proteins were broadly expressed across tissues, including the germline, showing an intracellular pattern characteristic of mitochondria based on MitoTracker staining (Suppl.Figure 1). However, we could not generate homozygous idha-1 null mutants using CRISPR/Cas9-mediated genomic deletion. While heterozygous animals showed mild developmental defects, their progeny homozygous for the idha-1 deletion mutation died as embryos or early larvae. Therefore, we conducted all subsequent experiments using RNAi to knock down idha-1.

Converted germ cells express additional neuronal genes

Because ectopic expression of a single reporter does not reflect a faithful change in cell identity, we tested whether converted germ cells activate neuronal genes beyond gcy-5, at both subtype-specific and pan-neuronal levels. For clarity, we refer to the CHE-1-induced germ cell conversion hereafter as GeCo (Germ Conversion).

We combined gcy-5^prom::GFP with RFP reporters for three additional identity levels: ceh-36^prom::RFP (the ASE/AWC-class homeobox gene acting downstream of che-1 in ASE specification), ift-20^prom::NLS::RFP (a sensory/ciliated neuron marker), and rab-3^prom::NLS::RFP (pan-neuronal). All three reporters were induced in converted germ lines of idha-1 RNAi animals (Figure 2A). Quantification showed that around 60% of animals with germline GFP also expressed either ceh-36^prom::RFP, rab-3^prom::RFP, or ift-20^prom::RFP (Figure 2B). Hence, CHE-1 induced GeCo co-activates ASE-specific, sensory, and pan-neuronal reporter gene expression, arguing that a broader neuronal program rather than a single promoter is being induced.

Figure 2. CHE-1-reprogrammed germ cells acquire neuronal characteristics.

(A) Representative images of ectopic neuronal reporter expression in the germline following idha-1 RNAi and broad CHE-1^OE. Animals expressed the ASE fate reporter gcy-5^prom::GFP with reporters for additional neuronal identity markers: ceh-36^prom::RFP, an ASE/AWC neuronal fate marker; ift-20^prom::NLS::RFP (marker of ciliated sensory neurons); and the pan-neuronal reporter rab-3^prom::NLS::RFP. For each reporter combination, merged DIC/GFP/RFP, GFP/RFP, and RFP channels are shown. Boxed regions indicate areas containing ectopic reporter-positive germ cells. These additional neuronal reporters were detected within gcy-5^prom::GFP positive germ cells following idha-1 depletion and CHE-1^OE. (B) Quantification of animals displaying co-expression of ectopic gcy-5^prom::GFP and the indicated neuronal RFP reporter in the germline following idha-1 RNAi and CHE-1^OE. Approximately 60% of animals showed co-expression of each neuronal reporter with gcy-5^prom::GFP, compared with ~3% in controls. **Data are presented as mean ± SEM. Statistical significance was determined by ordinary one-way ANOVA followed by Dunnett’s multiple comparisons test, with each condition compared with control. ***P < 0.001. Numbers within the bars indicate the total number of animals scored. (C) Detection of endogenous neuronal transcripts in reprogrammed germ cells by single-molecule fluorescence in situ hybridization (smFISH). Animals subjected to idha-1 RNAi and CHE-1^OE and displaying ectopic gcy-5 ^prom::GFP expression were hybridized with probes targeting endogenous gcy-5, rab-3, unc-119, or unc-10. GFP marks gcy-5 ^prom::GFP-positive germ cells; smFISH signals are shown as red puncta, and nuclei are counterstained with DAPI (blue). Boxed regions are shown at higher magnification in the bottom row. Endogenous transcripts for all four neuronal genes were detected within ectopic GFP-positive germ cells. (D) Immunofluorescence detection of the conserved presynaptic protein UNC-10/RIM in the germline. Representative control and idha-1 RNAi-treated animals following CHE-1^OE are shown. UNC-10 immunostaining is shown in red, gcy-5 ^prom::GFP in green, and DAPI in blue. Ectopic UNC-10 protein is detected within gcy-5 ^prom::GFP-positive germ cells following idha-1 depletion, whereas comparable neuronal marker expression is absent from the control germline. Boxed regions are shown at higher magnification. Short conclusion: Together, these results show that idha-1 depletion permits CHE-1 to activate a broader neuronal program in germ cells and supports the acquisition of a broader neuronal molecular identity rather than isolated activation of the gcy-5 fate reporter.

We confirmed that the transgenic reporters reflect genuine gene expression by examining endogenous transcripts using single-molecule fluorescence in situ hybridization (smFISH). In GFP-positive germlines of idha-1 RNAi animals, we detected smFISH signals for gcy-5 as well as the pan-neuronal genes rab-3, unc-119, and unc-10 (Figure 2C). Furthermore, antibody staining against the neuronal active-zone protein UNC-10/RIM revealed punctate UNC-10 signal within gcy-5^prom::GFP-positive germ cells of idha-1 RNAi animals, whereas control animals showed only the expected UNC-10 staining in the neighboring nerve cord and none in the germ line (Figure 2D).

Together, these data show that IDHA-1 depletion and ectopic CHE-1oe activate endogenous neuronal genes at the transcript and protein level in germ cells that acquire neuron-like morphology, corroborating the notion of genuine germ cell-to-neuron conversion.

We also asked whether the permissive state created by idha-1 depletion is specific to the ASE neuron fate. Ectopic expression of the Pitx-type homeodomain TF UNC-30, which specifies GABAergic motor neurons (Jin, Hoskins et al. 1994), induced the GABA transporter reporter unc-25^prom::GFP in the germline of idha-1 RNAi animals. However, ectopic expression of the myogenic bHLH factor HLH-1 induced no germline muscle reporter expression (Suppl.Figure 2). Hence, loss of IDHA-1 lowers a barrier that restricts conversion towards at least two distinct neuronal identities but does not seem to erase germ cell identity per se to allow induction of the muscle fate by HLH-1.

Loss of IDHA-1 reduces repressive histone methylation in the germline

Because αKG is the co-substrate of JmjC-domain histone demethylases, we asked whether depletion of IDHA-1 alters the histone methylation landscape. Western blotting of whole-worm lysates from idha-1 RNAi-treated animals for histone marks revealed essentially unchanged H3K4me3, which marks active chromatin and a modest reduction of the repressive histone marks H3K27me3 and H3K9me3 (Figure 3A). To validate IDHA-1 knockdown, we used the endogenous IDHA-1::3xHA allele for Western blotting of whole-worm lysates, with tubulin and 14-3-3 as loading controls (Figure 3A). Since whole-animal lysates contain mixed tissue including somatic and germ cells, we examined the germline directly by immunofluorescence, using HTZ-1/H2A.Z counterstaining for normalization (Figure 3B). Relative to control RNAi, idha-1 RNAi reduced germline H3K27me3 more pronouncedly, while H3K4me3 was slightly increased (Figure 3C). Thus, idha-1 knockdown reduces repressive histone marks, the opposite of what αKG limitation should cause if reduced αKG reduces JmjC activity that demethylates H3K9/27me3. Yet the observed loss of H3K9me3 and H3K27me3 upon idha-1 depletion explains the increased permissiveness for ectopic gene expression in germlines, as we observed in previous studies of barrier genes (Tursun, Patel et al. 2011, Patel, Tursun et al. 2012).

Figure 3. IDHA-1 depletion remodels histone methylation in the C. elegans germline.

(A) Whole-worm western blot analysis of histone methylation following RNAi-mediated depletion of idha-1. IDHA-1 depletion was confirmed in animals carrying endogenously tagged IDHA-1::3xHA using anti-HA antibodies. Whole-animal protein lysates from control and idha-1 RNAi-treated animals were probed for the active chromatin-associated histone modification H3K4me3 and the repressive histone modifications H3K27me3 and H3K9me3. α-tubulin/14-3-3 was used as loading controls. Depletion of idha-1 reduced H3K27me3 and H3K9me3 levels, indicating loss of repressive histone methylation following perturbation of mitochondrial IDH3. (B) Representative immunofluorescence images of germ lines from control and idha-1 RNAi-treated animals stained for H3K27me3, H3K9me3, or H3K4me3 (red). HTZ-1/H2A.Z staining (green) was used as a chromatin reference, and nuclei were counterstained with DAPI (blue). idha-1 depletion markedly reduced germline H3K27me3 and H3K9me3 fluorescence. Asterisks indicate the distal tip of the germline. Scale bars, 50 μm. (C) Quantification of relative fluorescence intensities of H3K27me3, H3K9me3, and H3K4me3 in germ lines from control and idha-1 RNAi-treated animals. idha-1 depletion markedly reduced the repressive histone modifications H3K27me3 and H3K9me3, accompanied by a significant increase in the active chromatin-associated mark H3K4me3. Data are presented as mean ± SEM. ****P < 0.0001; **P < 0.01.

This outcome suggests that chromatin state is not directly affected by reduced demethylase co-substrate availability and requires the genetic and metabolomic dissection of effects caused by idha-1 depletion described below.

GeCo requires HIF-1 and the JmjC-domain proteins JMJD-3.3 and JMJD-4

Knockdown of idha-1 is expected to interrupt the TCA cycle at the isocitrate-to-αKG step, which could create permissiveness for CHE-1-driven GeCo through several routes (Figure 4A). Besides JmjC oxygenase-mediate histone demthylation, another prominent αKG-dependent pathway is the hypoxia response: the prolyl hydroxylase EGL-9 (PHD) requires αKG to hydroxylate HIF-1 which then serves as signal for VHL-1-dependent ubiquitylation and proteasomal degradation (Figure 4B) (Epstein, Gleadle et al. 2001, Kaelin and Ratcliffe 2008).

Figure 4. HIF-1 and Jumonji histone demethylases promote IDHA-1 depletion-mediated germ cell reprogramming.

(A) Working model for mitochondrial regulation of transcription factor-induced germ cell reprogramming. Depletion of the mitochondrial TCA-cycle enzyme IDH3 through idha-1 RNAi is proposed to trigger retrograde effects from mitochondria to the nucleus, altering the chromatin environment and enabling ectopic CHE-1 to activate a neuronal gene program in germ cells. Question marks indicate unknown links between mitochondrial IDH3 depletion and nuclear changes. (B) Schematic of α-KG-dependent pathways linking IDH3 function to HIF-1 stability and chromatin regulation. The prolyl hydroxylase EGL-9/PHD uses α-KG for HIF-1 hydroxylation, enabling VHL-1-dependent ubiquitination and proteasomal degradation of HIF-1. α-KG is also a cofactor for Jumonji-domain oxygenases that catalyse histone demethylation. (C) Genetic modulation of HIF-1 signaling alters IDHA-1 depletion-mediated germ cell conversion (GeCo). Relative GeCo penetrance following idha-1 RNAi was quantified in control, hif-1^−/−, and vhl-1^−/− animals following CHE-1^OE. Loss of hif-1 strongly suppressed the idha-1-dependent reprogramming phenotype, whereas loss of vhl-1 enhanced GeCo, consistent with HIF-1 activity promoting reprogramming following IDH3 depletion. Statistical significance was determined by ordinary one-way ANOVA followed by Dunnett’s multiple-comparisons test, with mutant conditions compared with control. ****P < 0.0001. (D) RNAi suppressor screen for Jumonji-domain proteins required for IDHA-1 depletion-mediated reprogramming. Candidate Jumonji genes were co-depleted with idha-1, and GeCo penetrance following CHE-1^OE was normalized to the idha-1 RNAi control. Co-depletion of jmjd-3.3, jmjd-4, jmjd-5, jmjd-2, jmjd-3.2, jmjd-3.1, and jmjd-1.2 significantly suppressed the reprogramming phenotype, whereas depletion of jmjd-1.1, utx-1, psr-1, jmjc-1, or jhdm-1 did not significantly alter GeCo. The candidates jmjd-3.3 and jmjd-4 were selected for further validation. Statistical significance was determined by ordinary one-way ANOVA followed by Dunnett’s multiple-comparisons test, with each double-RNAi condition compared with the idha-1 RNAi control. ****P < 0.0001; **P < 0.01; ns, not significant. (E) Genetic interaction between HIF-1 and the selected candidate Jumonji factors jmjd-3.3 and jmjd-4. Stitched double-RNAi constructs were used to simultaneously deplete idha-1 and either jmjd-3.3 or jmjd-4, with idha-1::Rluc RNAi used as the corresponding control. In wild-type animals, co-depletion of either jmjd-3.3 or jmjd-4 significantly reduced GeCo compared with idha-1::Rluc RNAi. In the hif-1^−/− background, however, additional depletion of jmjd-3.3 or jmjd-4 did not further suppress the reprogramming phenotype, consistent with HIF-1 and these Jumonji factors acting within the same genetic pathway. Data are presented as mean ± SEM. Numbers within bars indicate the total number of animals scored. Statistical significance was determined by ordinary one way ANOVA followed by Tukey’s multiple comparisons test. ****P < 0.0001; ns, not significant.

Using a viable hif-1 null mutant background, we observed that GeCo upon idha-1 RNAi was significantly reduced by approximately 50%. Conversely, vhl-1 null animals, in which HIF-1 protein is constitutively stabilized (Bishop, Lau et al. 2004, Shen, Nettleton et al. 2005), showed strong GeCo enhancement by 1.5-fold (Figure 4C). HIF-1 activity appears to be rate-limiting but also to promote GeCo upon idha-1 depletion. This effect is not caused by altered levels of the reprogramming-driving TF CHE-1, as Western blotting of heat-shocked animals carrying CHE-1::3xHA showed equal levels of induced CHE-1 protein in wild-type and hif-1 mutant backgrounds (Suppl.Figure 3A).

The observation of reduced histone methylation (Figure 3B,C) suggested that αKG-dependent dioxygenases that mediate histone demethylation may be required for CHE-1-induced GeCo. We screened the C. elegans JmjC-domain gene family by double RNAi combined with idha-1 knockdown and found that co-depletion of jmjd-3.3 and jmjd-4 most strongly suppressed GeCo, while a second tier of genes, including jmjd-5, jmjd-2, jmjd-3.2, jmjd-3.1, and jmjd-1.2, gave statistically significant but weaker suppression (Figure 4F). The available mutant alleles for jmjd-3.3 and jmjd-4 showed weaker suppression than RNAi (Suppl. Figure 3B), which we attribute to genetic compensation within this partially redundant family; the two available jmjd-3.3 alleles are short in-frame deletions that may preserve at least partial protein activity. We therefore followed up on jmjd-3.3 and jmjd-4 with stitched double-RNAi clones, which can be used for more stable double knockdown assessments (Kazmierczak, Farre et al. 2021). The idha-1 RNAi clone fragment is fused in the L4440 RNAi vector with a jmjd-3.3 or jmjd-4 genomic fragment, or with the Rluc gene fragment as the matched control. Stitched idha-1::jmjd-3.3 reduced GeCo by around 25%, confirming the single-clone result (Figure 4G). Notably, when the same stitched double RNAi was performed in the hif-1 null background, GeCo was reduced by idha-1::Rluc compared to the wild-type background but not further suppressed by co-depletion of jmjd-3.3 or jmjd-4 relative to idha-1::Rluc in the hif-1 null background (Figure 4G). Hence, the absence of an additive effect places HIF-1 and these two JmjC proteins in a common genetic pathway for GeCo rather than in parallel pathways. Overall, these findings identify HIF-1 and the JMJDs as central players in creating permissiveness for DR of germ cells in the idha-1 depletion context. Yet, the directionality of jmjd-3.3 and jmjd-4 depletion and their effect on chromatin is not concordant, as reduced demethylase activity should increase histone methylation. This suggests additional or alternative pathways affected by IDH3 activity. Also, because GeCo is not fully abolished upon loss of HIF-1 or JMJD proteins, further effects may be implicated in licensing DR of germ cells to neurons upon idha-1 depletion.

Metabolic consequences of idha-1 depletion

If the permissive state resulted from reduced αKG production, supplying αKG to the idha-1-depleted worms should suppress GeCo. We therefore fed animals defined TCA cycle intermediates combined with idha-1 RNAi and scored GeCo (Figure 5A). Unexpectedly, αKG feeding produced the opposite effect, causing a small enhancement, while malate and fumarate enhanced GeCo by up to 1.5-fold (Figure 5B). Succinate and citrate feeding had no significant effect at the tested concentrations of 10 and 15 mM in the growth medium (Figure 5B). Thus, exogenous supply of the TCA intermediates αKG, malate, and fumarate modulates conversion, but not in the direction that αKG-scarcity would predict.

Figure 5. IDHA-1 depletion induces metabolic rewiring.

(A) Schematic of the metabolite-feeding assay. Animals were treated with idha-1 RNAi and supplemented with the indicated TCA-cycle metabolites before broad CHE-1^OE. Germ cell conversion was assessed by ectopic activation of the ASE fate reporter gcy-5^prom::GFP. (B) Effect of exogenous TCA-cycle metabolites on IDHA-1 depletion-mediated germ cell reprogramming. Animals treated with idha-1 RNAi were supplemented with α-ketoglutarate (α-KG), succinate, fumarate, malate or citrate at the indicated concentrations, and GeCo penetrance following CHE-1^OE was quantified relative to the vehicle-treated idha-1 RNAi control. α-KG modestly enhanced reprogramming, whereas fumarate and malate produced a stronger increase in GeCo. In contrast, succinate and citrate did not significantly alter the phenotype. Data are presented as mean ± SEM. Statistical significance was determined by one-way ANOVA followed by Dunnett’s multiple comparisons test. **P < 0.01; *P < 0.05; ns, not significant. (C) Schematic of the stable isotope tracing experiment used to determine whether carbon flow through the TCA cycle is maintained following IDH3 depletion. F1 animals were exposed to control or idha-1 RNAi throughout development and subsequently fed ¹³C-labelled glucose. Animals were collected after 60, 180, 360, and 840 min of isotope labeling, followed by metabolite extraction and gas chromatography-mass spectrometry (GC-MS). Time-dependent incorporation of ¹³C into the TCA cycle and associated metabolites was used to assess relative changes in carbon utilization and metabolic pathway activity following idha-1 depletion. (D) Time-resolved ¹³C glucose tracing of TCA-cycle and glutamine-associated metabolites following control (Rluc) or idha-1 RNAi. Despite depletion of the TCA-cycle enzyme IDHA-1, ¹³C incorporation into α-KG, fumarate, and malate remained broadly comparable with control, indicating that downstream TCA-cycle metabolism is at least partially maintained. Citrate showed increased ¹³C labeling following idha-1 depletion, consistent with accumulation of carbon upstream of the disrupted IDH3-dependent isocitrate-to-α-KG step. Succinate and glutamine showed modest decreases, while glutamate showed a more pronounced reduction in ¹³C labeling following IDH3 depletion, indicating increased utilization of glutamate-associated carbon. Individual symbols represent the 6 independent biological replicates.

We therefore performed metabolomics to directly measure the metabolic consequences of idha-1 depletion. Using stable-isotope-resolved metabolomics, we exposed P0 L4 animals to idha-1 or control (Rluc) RNAi and maintained their F1 progeny on RNAi until the L4 stage, when animals were fed ¹³C-glucose for 60, 80, 260, or 840 minutes before extraction and GC-MS analysis, with six biological replicates per condition and time point (Figure 5C). The dominant effect was a strong accumulation of citrate in idha-1-depleted animals, apparent at 60 minutes and increasing over the time course (Figure 5D), as expected when blocking the isocitrate dehydrogenase step. Unexpectedly, changes downstream of the block were modest, with αKG showing no significant difference between idha-1 and control RNAi at any time point. Only succinate was modestly reduced at 260 and 840 minutes. However, glutamate was consistently reduced in idha-1 RNAi (Figure 5D). An independent metabolomics experiment without isotope labeling, comparing P0 and F1 idha-1 RNAi animals at a single L4 time point, reproduced this pattern, including citrate accumulation and the glutamate decrease (Suppl. Figure 4). These results suggest that GeCo upon idha-1 depletion occurs without reducing the whole-animal steady-state αKG pool. Yet, this does not exclude a local or compartment-specific change, because GC-MS of whole-worm extracts averages across tissues, and therefore we also attempted GC-MS of dissected gonads from idha-1 RNAi animals. However, using around 100 dissected germlines as one sample did not provide enough material for even one proper GC-MS run.

Nevertheless, our GC-MS analysis rules out bulk αKG scarcity and is consistent with the observation that αKG feeding fails to suppress conversion. Additionally, the reduction in glutamate indicated altered flux through the glutamine/glutamate node, which prompted us to test this branch genetically.

Glutamate anaplerosis modulates germ cell conversion

Glutamine and glutamate pools can replenish the TCA cycle at the αKG node by glutaminases (GLNA-1, GLNA-3) hydrolyzing glutamine to glutamate, and by glutamate dehydrogenase (GDH-1) converting glutamate to αKG (Owen, Kalhan et al. 2002, Yang, Ko et al. 2014). Conversely, glutamine synthetase GLN-2 runs the opposite reaction, and αKG can additionally be generated from α-hydroxyglutarate by the hydroxyglutarate dehydrogenases DHGD-1/LHGD-1 (Figure 6A). Testing these possible pathways by double RNAi in the idha-1 RNAi background showed that glutamine synthetase knockdown by targeting gln-2, which is predicted to increase the glutamate pool available for anaplerosis, enhanced GeCo nearly 2-fold (Figure 6B). In contrast, knockdown of the glutaminases GLNA-1 and GLNA-3, as well as of glutamate dehydrogenase GDH-1, strongly suppressed DR of germ cells into neurons. We generated a glna-3 null mutant which confirms the RNAi phenotype. (Supplemental Figure 5). RNAi against dhgd-1 or lhgd-1 had no significant effect (Figure 6B), indicating that the anaplerotic flux from glutamine through glutamate into αKG is required for GeCo. Hence, shifting the balance of this anaplerosis node towards glutamate by removing glutamine synthetase GLN-2 makes germ cells significantly more permissive.

Figure 6. Glutamine anaplerosis and downstream TCA cycle activity support IDH3 depletion-mediated germ cell reprogramming.

(A) Schematic of metabolic pathways capable of maintaining downstream TCA cycle metabolism following IDH3 depletion. IDH3 normally catalyzes the conversion of isocitrate to α-KG. Alternative α-KG-generating routes include conversion of glutamine to glutamate by the glutaminases GLNA-1 and GLNA-3, followed by conversion of glutamate to α-KG by GDH-1, as well as hydroxyglutarate metabolism. GLN-2 catalyzes the opposing conversion of glutamate to glutamine. The ICL-1-dependent glyoxylate shunt represents an alternative route that could, in principle, maintain TCA cycle function by bypassing the IDH3-dependent isocitrate to α-KG step. Enzymes from these pathways and the downstream TCA cycle were therefore examined genetically. (B) Genetic analysis of alternative α-KG generating pathways during IDHA-1 depletion-mediated reprogramming. Candidate metabolic enzymes were co-depleted with idha-1, and GeCo penetrance following CHE-1^OE was quantified relative to the idha-1 RNAi control. Depletion of the glutaminases glna-1 and glna-3, or glutamate dehydrogenase gdh-1, suppressed reprogramming, whereas depletion of glutamine synthetase gln-2, which directs glutamate back toward glutamine, enhanced GeCo. In contrast, depletion of the hydroxyglutarate dehydrogenases dhgd-1 and lhgd-1 did not significantly alter the phenotype. These genetic interactions support a functional contribution of glutamate-to-α-KG metabolism during IDH3 depletion. (C) Requirement for downstream TCA cycle metabolism during IDHA-1 depletion-mediated reprogramming. TCA cycle and glyoxylate shunt enzymes were co-depleted with idha-1, and GeCo was quantified relative to the idha-1 RNAi control. Co-depletion of ogdh-1, sdhb-1, or fum-1, which function downstream of α-KG, significantly suppressed reprogramming, whereas depletion of aco-2, mdh-1, or the glyoxylate-shunt enzyme icl-1 did not significantly alter the phenotype. These results indicate that metabolism downstream of α-KG is required for efficient IDH3 depletion-mediated reprogramming, while the ICL-1-dependent glyoxylate shunt is dispensable. (D) Working model integrating metabolic and HIF-1-dependent regulation of IDHA-1 depletion-mediated germ cell reprogramming. IDH3 depletion establishes a reprogramming-permissive state that enables CHE-1^OE to induce neuronal conversion. **Data are presented as mean ± SEM. Statistical significance was determined by ordinary one-way ANOVA followed by Dunnett’s multiple-comparisons test, with each double-RNAi condition compared with the idha-1 RNAi control. ***P < 0.001; **P < 0.01; *P < 0.05; ns, not significant.

Consistent with a requirement for glutamate handling rather than for glutamine itself, glutamine feeding had no significant effect on GeCo, whereas RNAi against the glutamate transporter GLFT-1 reduced conversion (Supplemental Figure 5).

We also tested whether depletion of TCA enzymes acting downstream of the idha-1 block affected GeCo. Depletion of the αKG dehydrogenase OGDH-1, the succinate dehydrogenase subunit SDHB-1, and the fumarase FUM-1 each strongly suppressed GeCo, whereas depletion of the aconitase ACO-2, the malate dehydrogenase MDH-1, and the isocitrate lyase ICL-1 had no significant effect (Figure 6D). Thus, GeCo requires continued flux from αKG to malate and is triggered by loss of flux at the isocitrate-to-αKG step. This is consistent with the metabolite-feeding data, in which fumarate and malate enhanced rather than suppressed conversion.

Taken together, these results reveal two conditions caused by idha-1 depletion that contribute to the permissiveness of germ cells DR to neurons: elevated HIF-1 and glutamate anaplerosis levels (Figure 6E).

The putative glucose transporter FDGT-2 is an additional GeCo barrier

To systematically search for additional metabolic modifiers, we performed an extended double RNAi screen with idha-1 RNAi combined with ~110 genes encoding metabolic enzymes, solute carriers, and transporters, including the glutamate anaplerosis and TCA genes tested above as internal controls (Figure 7A). The screen reproduced these and also identified a new target: RNAi against fdgt-2, which encodes a facilitated glucose transporter of the GLUT/SLC2A family (Kim, Underwood et al. 2018), markedly increased conversion relative to idha-1 RNAi alone (Figure 7A). We confirmed this finding with a CRISPR-generated fdgt-2 null mutant, in which GeCo upon idha-1 RNAi was enhanced almost 2-fold over idha-1 depletion alone (Figure 7B). We also generated an FDGT-2::3xHA knock-in allele to assess expression and could detect broad expression across tissues (Supplemental Figure 6).

Figure 7. Genetic modifier screening identifies transporter FDGT-2 as a restraint on IDH3 depletion-mediated GeCo.

(A) Summary of genetic modifiers of IDHA-1 depletion-mediated germ cell reprogramming identified by secondary RNAi screening. Candidate genes were co-depleted with idha-1, and CHE-1^OE-induced GeCo was quantified relative to idha-1 RNAi alone. Values are normalized to the idha-1 RNAi condition, indicated by the dashed line at 1; values below or above this reference indicate suppression or enhancement of reprogramming, respectively. The putative monosaccharide transporter fdgt-2 is highlighted in red, while selected genes involved in glutamine/α-KG metabolism are highlighted in green. The screen identified both suppressors and enhancers of the idha-1-dependent phenotype, with depletion of fdgt-2 producing a prominent enhancement of reprogramming. Data are presented as mean ± SEM. Statistical significance was determined using ordinary one-way ANOVA followed by Dunnett’s multiple-comparisons test, comparing each co-depletion condition with the idha-1 RNAi control. ***P < 0.001; ns, not significant. (B) Genetic validation of FDGT-2 as a negative regulator of IDHA-1 depletion-mediated reprogramming. GeCo following idha-1 RNAi was quantified in control and fdgt-2−/− (CRISPR deletion) animals and normalized to the control response. Loss of fdgt-2 significantly enhanced reprogramming following IDHA-1 depletion. Data are presented as mean ± SEM. Statistical significance was determined using an unpaired two-tailed Student’s t-test. **P < 0.01. (C) Integrated model for metabolic regulation of IDH3 depletion-mediated germ cell reprogramming. Loss of mitochondrial IDH3 disrupts the canonical isocitrate-to-α-KG step of the TCA cycle and engages multiple compensatory and signaling responses, including HIF-1 activity, Jumonji-domain histone demethylases, glutamine anaplerosis, and altered monosaccharide transport. HIF-1-dependent signaling, Jumonji activity, and metabolic rerouting toward α-KG are proposed to converge with the observed changes in histone methylation to establish a reprogramming-permissive state. In contrast, FDGT-2-dependent transport provides a compensatory metabolic response that limits the extent of reprogramming. Together, these pathways determine whether differentiated germ cells remain refractory to CHE-1 or acquire competence for neuronal conversion.

This finding indicates that reduced glucose uptake acts in the same direction as elevated HIF-1 and reduced GLN-2, adding a third metabolic input caused by idha-1 depletion that increases germ-cell permissiveness to TF-induced conversion. Together, our results support a model in which perturbation of the TCA cycle at the isocitrate-to-αKG step causes HIF-1 stabilization, JmjC-dependent activities, altered glutamate anaplerosis, and reduced glucose transport, which converge on the germ cell nucleus to diminish the repressive H3K9me3 and H3K27me3 histone modifications, allowing ectopic CHE-1 to activate neuronal gene expression resulting in DR for germ cells into neuron-like cells (Figure 7C).

DISCUSSION

The most direct interpretation of an IDH3 block causing less αKG, and therefore less activity of αKG-dependent dioxygenases such as JmjC demethylases, resulting in maintained or even more repressive histone methylation is contradicted by our finding. Germline H3K9me3 and H3K27me3 decrease rather than increase, whole-animal steady-state αKG is not measurably reduced by idha-1 RNAi based on our different metabolomics measurements, and feeding αKG does not suppress conversion. These apparent contradictions are informative rather than merely negative, and they suggest several possible explanations.

We note that GC-MS of whole-worm extracts reports a population average across tissues, and more tissue-resolved information on αKG would provide additional hints towards the causative effects. To test this, we would need to measure the relevant αKG levels for JmjC and PHD enzymes in germ cells by metabolomics on isolated gonads. Our attempt at GC-MS using 100 isolated gonads per sample was unsuccessful, and a direct αKG sensor is currently unavailable. A potential sensor based on the ATP synthase ATP-2 could be an alternative route (Chin, Fu et al. 2014), or collecting more gonads. But establishing these approaches is beyond the scope of the present study.

Notably, αKG-dependent dioxygenases are competitively inhibited by succinate and fumarate, so their activity depends on the αKG/succinate and αKG/fumarate ratios rather than on absolute αKG alone (Xiao, Yang et al. 2012, TeSlaa, Chaikovsky et al. 2016). Succinate is modestly reduced at later time points upon idha-1 RNAi while αKG is unchanged, which would raise the αKG/succinate ratio and thus increase dioxygenase activity, consistent with the loss of repressive marks we observe, and with the requirement for JMJD-3.3 and JMJD-4. Yet, the enhancement of GeCo by fumarate and malate feeding does not fit this framework and may instead reflect anaplerotic replenishment of the distal cycle, consistent with the requirement for ogdh-1, sdhb-1, and fum-1. Yet, the requirement for jmjd-4 is interesting, because the mammalian orthologue JMJD4 is not a histone demethylase. JMJD4 is an αKG-dependent lysyl hydroxylase that hydroxylates the translation termination factor eRF1 at K63 and is required for optimal translational termination (Feng, Yamamoto et al. 2014, Zhuang, Feng et al. 2015). If the C. elegans protein retains this activity, JMJD-4 may contribute to GeCo through protein hydroxylation and translational control rather than chromatin.

HIF-1 stabilization by eliminating VHL-1 increases GeCo, while HIF-1 loss decreases it. The fact that several JmjC demethylases are HIF-1 target genes in other systems (Pollard, Loenarz et al. 2008, Wellmann, Bettkober et al. 2008, Krieg, Rankin et al. 2010) provides a plausible feedforward route by which even a modest change in PHD activity based on αKG levels is translated into a larger change in demethylase dosage. While our study establishes a genetic requirement chain from IDH3 activity to HIF-1, JMJD-3.3/JMJD-4, and permissive chromatin required for GeCo, dissecting the exact metabolic link will require tissue-specific chromatin profiling in germ cells combined with direct measurement of the relevant metabolite pools. These insights are central goals of future work rather than gaps to close in the present study.

Furthermore, we provide evidence that GeCo requires flux through glutamate into αKG upon idha-1 RNAi, which is supported by metabolomics showing that glutamate is significantly reduced upon idha-1 RNAi. Glutamine is rather unchanged, indicating increased glutamate consumption rather than a shortage of its precursor, and that the glutamine pool may be buffered by dietary uptake or alternative synthesis routes. Hence, when the canonical route to αKG is blocked at IDH3, cells appear to rely on glutamate anaplerosis as previously shown in other systems (Owen, Kalhan et al. 2002, DeBerardinis, Mancuso et al. 2007, Xiao, Zeng et al. 2016). In particular, the TCA block in cancer cells increases glutamate dehydrogenase (GDH) activity to generate αKG from glutamate, consistent with our finding that GDH-1 is required for GeCo upon TCA block by IDHA-1 depletion (Yang, Ko et al. 2014). Overall, manipulations that increase or decrease the glutamate available for this route in C. elegans enhance or suppress DR of germ cells to neurons accordingly.

The targeted RNAi screen for additional metabolic factors identified fdgt-2, which encodes a facilitated glucose transporter of the GLUT/SLC2A family (Kim, Underwood et al. 2018), as another strong enhancer. This finding indicated that reduced glucose uptake may limit glycolytic flux and pyruvate supply to the cycle, compounding the IDH3 block. Notably, loss of FDGT-2 at the glucose-uptake step would be expected to increase reliance on glutamate anaplerosis as well (Yang, Ko et al. 2014). While the exact function of FDGT-2 has not yet been studied experimentally, its identification supports the broader point of this study: multiple metabolic perturbations based on impaired αKG production, including HIF-1 stabilization, altered glutamate anaplerosis, and potentially reduced glucose import, shift germ cells towards permissiveness for DR to neuron-like cells.

The list of characterized reprogramming barriers so far is dominated by chromatin factors such as LIN-53/RBBP4–7 (Tursun, Patel et al. 2011), CAF1 (Cheloufi, Elling et al. 2015), PRC2 (Patel, Tursun et al. 2012), FACT (Kolundzic, Ofenbauer et al. 2018), MRG-1 (Hajduskova, Baytek et al. 2019), and more. Although a metabolic barrier joins this list, epigenetic regulation remains the ultimate mechanism for maintaining germ cell identity through repressive H3K27me3 and H3K9me3 chromatin domains. Our data indicate that this state is metabolically supported and requires IDH3 activity. That conversion proceeds to two neuronal fates (ASE via CHE-1, GABAergic via UNC-30) but not to a muscle fate under the same conditions suggests that removing the metabolic barrier creates sufficient permissiveness only for the neuronal program, potentially because the MyoD homolog HLH-1 alone is not strong enough to induce the muscle fate in the context of reduced metabolic barrier. This result mirrors the fate-selectivity previously observed for chromatin barriers, where FACT depletion permits different conversions in different tissues (Kolundzic, Ofenbauer et al. 2018).

Beyond direct reprogramming, our findings are also related to IDH pathology in humans. Certain IDH1/IDH2 mutations cause excessive production of the oncometabolite 2-hydroxyglutarate, which inhibits αKG-dependent dioxygenases and increases repressive histone methylation, thereby blocking differentiation in glioma and leukemia (Figueroa, Abdel-Wahab et al. 2010, Xu, Yang et al. 2011, Lu, Ward et al. 2012). In the C. elegans germline, loss of the NAD-dependent isozyme IDH3 has the opposite consequence since it is required to maintain repressive histone methylation. That two different isocitrate dehydrogenases can push chromatin and cell-fate plasticity in opposite directions underscores how much cell identity is negotiated at this metabolic node of the TCA flux.