CLUH controls astrin-1 expression to couple mitochondrial metabolism to cell cycle progression

Proliferating cells undergo metabolic changes in synchrony with cell cycle progression and cell division. Mitochondria provide fuel, metabolites, and ATP during different phases of the cell cycle, however it is not completely understood how mitochondrial function and the cell cycle are coordinated. CLUH (clustered mitochondria homolog) is a post-transcriptional regulator of mRNAs encoding mitochondrial proteins involved in oxidative phosphorylation and several metabolic pathways. Here, we show a role of CLUH in regulating the expression of astrin, which is involved in metaphase to anaphase progression, centrosome integrity, and mTORC1 inhibition. We find that CLUH binds both the SPAG5 mRNA and its product astrin, and controls the synthesis and the stability of the full-length astrin-1 isoform. We show that CLUH interacts with astrin-1 specifically during interphase. Astrin-depleted cells show mTORC1 hyperactivation and enhanced anabolism. On the other hand, cells lacking CLUH show decreased astrin levels and increased mTORC1 signaling, but cannot sustain anaplerotic and anabolic pathways. In absence of CLUH, cells fail to grow during G1, and progress faster through the cell cycle, indicating dysregulated matching of growth, metabolism, and cell cycling. Our data reveal a role of CLUH in coupling growth signaling pathways and mitochondrial metabolism with cell cycle progression.


Introduction
Proliferating cells adjust their metabolism to each individual phase of the cell cycle (DeBerardinis et al, 2008;Buchakjian & Kornbluth, 2010;Lee & Finkel, 2013;Salazar-Roa & Malumbres, 2017). Before division cells must double their biomass and DNA content. This entails a prominent rewiring of metabolism to foster biosynthetic pathways allowing the synthesis of nucleic acids, proteins and lipids. During mitosis, an increase in ATP production supplies the energy needed for pulling apart and segregating the chromosomes. As metabolic organelles, mitochondria are key actors in the metabolic reprogramming of proliferating cells.
On the one hand, cell cycle regulators influence mitochondrial metabolism in various ways (Leal-Esteban & Fajas, 2020). For example, CDK1 phosphorylates components of the respiratory chain during G2 and of the mitochondrial translocase of the outer membrane during mitosis, probably to enhance oxidative respiration and ATP production (Harbauer et al, 2014;Wang et al, 2014). On the other hand, energy deficiency is sensed by cell cycle checkpoints causing cell cycle arrest (Mandal et al, 2005;Salazar-Roa & Malumbres, 2017). Mitochondrial dynamics is also intimately linked to cell cycle progression. A hyperfused mitochondrial network is important for the G1-S transition, while mitochondria must fragment before cells enter mitosis (Mitra et al, 2009). Therefore, proliferating cells depend on several coordinated mechanisms to regulate how mitochondria utilize nutrients to fuel cell growth or energy production during the cell cycle. However, how mitochondrial metabolism and respiratory function are coordinated during the cell cycle is still largely unclear. Unravelling these mechanisms is crucial since they are often dysregulated in cancer.
One of the ways how mitochondrial metabolism can be rewired is through regulation of RNA regulons by RNA-binding proteins (Schatton & Rugarli, 2018). CLUH (clustered mitochondria homologue) is a conserved RNA-binding protein, which binds transcripts encoding proteins involved in the respiratory chain, the TCA cycle, and other mitochondrial metabolic pathways (Gao et al, 2014). Upon loss of CLUH, target mRNAs are subjected to faster decay and their respective proteins are decreased in abundance (Schatton et al, 2017). This leads to alterations in mitochondrial distribution (mitochondrial clustering as the gene name alludes to) mitochondrial cristae integrity, respiratory defects, loss of mtDNA, and decreased activity of TCA cycle enzymes (Schatton et al, 2017;Wakim et al, 2017;Pla-Martin et al, 2020). In the mouse, constitutive loss of Cluh causes neonatal lethality, while a liver-specific Cluh knockout (KO) led to hypoglycemia and defective ketogenesis upon starvation (Schatton et al, 2017).
CLUH and its target mRNAs form ribonucleoprotein particles in primary hepatocytes. These CLUH granules are important not only to protect target mRNAs from degradation, but also to restrict mTORC1 activation and promote mitophagy (Pla-Martin et al, 2020). Thus, in postmitotic hepatocytes CLUH controls the balance between catabolism and anabolism, allowing a survival response to starvation.
Astrin (also called SPAG5) is a multifunctional protein with a well-established role during mitosis. The complex astrin/kinastrin/DYNLL1 is involved in the stabilization of the kinetochore-microtubule interactions in metaphase (Kern et al, 2017). Consistently, cells lacking astrin or its interactor kinastrin show defects in chromosome segregations and a delayed metaphase to anaphase progression (Gruber et al, 2002;Dunsch et al, 2011). Astrin has however other roles in interphase, ranging from mTORC1 inhibition (Thedieck et al, 2013), centrosome integrity and centriole duplication (Thein et al, 2007;Kodani et al, 2015), and regulation of p53 levels during G2 after DNA damage (Halim et al, 2013). Astrin is highly expressed in several cancers and it is considered a poor prognostic marker (Yuan et al, 2014;Abdel-Fatah et al, 2016;Bertucci et al, 2016;Zhou et al, 2018;Ying et al, 2020).
We show here that in HeLa cells astrin exists in two isoforms, astrin-1 and astrin-2. We reveal that CLUH binds the SPAG5 mRNA, as well as the astrin-1 protein. The interaction of CLUH and astrin-1 occurs during interphase and protects astrin-1 and kinastrin from degradation.
Moreover, we show that CLUH positively regulates mitochondrial anaplerotic pathways that sustain mTORC1 activation. Our data disclose a role of CLUH in coupling mitochondrial metabolism to cell cycle progression.
Astrin appears as two isoforms (hereafter referred to as astrin-1 and astrin-2) in western blots, and CLUH interacts preferentially with the slower migrating isoform, astrin-1 ( Figure 1A). The existence of these two isoforms has been reported in the literature but their identity remains enigmatic. On closer inspection of the SPAG5 mRNA, we noticed the presence of five in-frame ATGs downstream of the first ATG (hereafter referred to as ATG2-6), which could serve as alternative translation start codons (with moderate and good Kozak sequences) ( Figure 1C).
Moreover, an upstream ORF within the SPAG5 5' UTR ( Figure 1C) could influence translation from the downstream ATGs as reported for other genes (Morris & Geballe, 2000;Young & Wek, 2016). To test these hypotheses, we produced different constructs expressing the 5' UTR and the coding region of the SPAG5 cDNA, and we mutagenized the uATG, the ATG1 or downstream ATGs individually or in different combinations to GGG (Figure 1-figure   supplement 1A). When we expressed the uATG GGG -SPAG5 construct, the expression of astrin-2 was suppressed ( Figure 1D). Furthermore, mutagenesis of ATG1 confirmed the existence of a downstream alternative translation start site ( Figure 1D). Overexpression of the constructs harboring mutagenized downstream ATGs revealed that astrin-2 is the product of alternative translation starting from ATG3-5 ( Figure 1E). The identity of astrin-1 was also confirmed by pull-down of this isoform only using an antibody directed against the N-terminal region of the protein (Figure 1-figure supplement 1B).
To evaluate whether the interaction domain of astrin that binds to CLUH lies in the N-terminal part of astrin-1, we transfected a construct expressing both astrin-1 and astrin-2 (ATG1-SPAG5,  Figure 1F). An interaction was observed only with full-length astrin, proving that CLUH binds within the N-terminus of astrin-1 ( Figure 1F). In contrast, the astrin interactor kinastrin has no binding preference to any of the two astrin isoforms, consistent with the previously mapped interaction domain between these two proteins (residues 482-693 of astrin) (Figure 1-figure supplement 1C) (Dunsch et al, 2011;Friese et al, 2016;Kern et al, 2016;Kern et al, 2017). CLUH still interacts with astrin-1 upon kinastrin depletion, suggesting that kinastrin is not required for the interaction (Figure 1-figure supplement 1D). Overexpression of both astrin-1 and astrin-2 led to a diffuse cytosolic staining and to the formation of granular structures that vary in size, likely depending on the level of expression. Endogenous CLUH was characteristically recruited to astrin-1 but not astrin-2 positive structures, in agreement with the biochemical characterization of this interaction ( Figure 1G).
In polysome profiles performed after chemical crosslinking, CLUH migrated in lighter fractions, as well as in fractions containing the monosome and polysomes, while astrin and kinastrin were only detected in the lighter fractions ( Figure 1H), indicating the existence of distinct CLUH complexes. In conclusion, we identify the expression of two astrin isoforms via alternative initiation of translation, and establish CLUH as a specific molecular partner of astrin-1.

CLUH controls astrin-1 and kinastrin stability and astrin expression
In absence of CLUH, kinastrin and astrin-1, but not astrin-2, are decreased in abundance ( Figure   2A). To investigate whether CLUH influences the stability of its interaction partners, we performed a cycloheximide (CHX) chase in wildtype and CLUH KO HeLa cells (Figure 2A).
The underlying reason for the only partial rescue of astrin-1 might be another level of regulation by CLUH. Given the identification of SPAG5 mRNA as a CLUH target (Gao et al, 2014), we tested if CLUH is involved in astrin synthesis. To this end, we immunoprecipitated newly synthesized astrin in wildtype and CLUH KO cells after labeling with 35 S methionine for different time points ( Figure 2H-I). We observed that newly synthesized astrin-1 is reduced in absence of CLUH to a greater extent than astrin-2, possibly also reflecting the decreased stability of this isoform ( Figure 2H-I). CLUH controls the expression of some of its target mRNAs by preventing their degradation (Schatton et al., 2017). To elucidate whether this is also the case for SPAG5, we labeled newly synthesized RNA with the uridine analog ethynyl uridine (EU) and chased it for 8 h. We precipitated the EU-RNA with biotin-azide using Click-iT chemistry and quantified SPAG5 mRNA levels with qRT-PCR. SPAG5 mRNA was less stable in absence of CLUH, whereas steady state levels and EU incorporation were unaffected ( Figure 2J-L). In conclusion, our data reveal that CLUH controls the expression of astrin-1 at multiple levels, by binding and stabilizing its mRNA, by affecting its synthesis, and maintaining its stability.

CLUH binds astrin-1 in interphase
Astrin controls cell cycle progression at various stages, raising the question if the complex CLUH-astrin-1 is cell cycle-dependent. To this end, we immunoprecipitated CLUH from cells enriched in different phases of the cell cycle ( Figure 3A, Figure 3-supplement 1A). The effective synchronization of the cells was confirmed by analyzing the input levels of cyclin D3 (which increases in G1 and G2), pH3-Ser10 (M phase marker) and pRPS6-Ser235/236 (inhibited upon starvation) ( Figure 3A). Astrin-1, but not astrin-2, was more abundant in S and G2, while CLUH levels were similar in the different phases of the cell cycle ( Figure 3A). Both astrin isoforms are phosphorylated at multiple sites, leading to slower migration in SDS-PAGE in prometaphase (PM) ( Figure 3A) (Chang et al, 2001;Cheng et al, 2008;Chiu et al, 2014;Chung et al, 2016;Geraghty et al, 2021). The CLUH/astrin-1 complex is enriched in S and G2 phases but not detected in PM, indicating that CLUH preferentially interacts with unmodified astrin-1 in interphase and that the interaction might be controlled by the phosphorylation status of astrin ( Figure 3A).
To fully capture the CLUH interactome, we performed four biological independent IPs of CLUH in G2-synchronized wildtype cells followed by label-free mass spectrometry ( Figure   3B, Figure 3-figure supplement 1A). The correct synchronization was confirmed by propidium iodide (PI) staining followed by flow cytometry (Figure 3-figure supplement 1B, C). This experiment confirmed astrin-1 and kinastrin as highly enriched interactors of CLUH ( Figure   3B and Supplementary File 2). In addition, proteins previously shown to be part of the astrin/kinastrin complex, such as MYCBP, or linking the complex to the microtubule cytoskeleton were also detected in the precipitate. Among these proteins are CLASP1 and 2, two microtubule plus-end binding proteins, and DYNLL1, a retrograde motor protein (Manning et al, 2010;Schmidt et al, 2010;Dunsch et al, 2011;Kern et al, 2016;Kern et al, 2017) ( Figure   3B). Interestingly, among top CLUH interactors, we found the centrosomal proteins CEP170B and CEP44, suggesting a link of CLUH with the centrosome and the microtubules ( Figure 3B, Since astrin has been previously found at the centrosome and pericentriolar material (Cheng et al, 2007;Thein et al, 2007;Kodani et al, 2015), we explored if astrin-1 resides in these organelles, using respective markers, such as CEP43 and PCM1. To avoid analyzing cells with large astrin-1 aggregates, we focused on cells expressing lower levels of astrin-1. In some cells, astrin-1 labelled a dotty structure that was positive at one extremity for CEP43 and partially colocalized with PCM1 ( Figure 3D). However, in other cells astrin-1-positive punctae were not decorated by centrosomal markers ( Figure 3D). This suggests that overexpressed astrin-1 can accumulate at the centrosome, but also at non centrosomal structures. Since lack of specific antibodies against endogenous astrin-1 precluded us to further explore this localization, we asked if endogenous CLUH can be found at the centrosome. CLUH and CEP43 stainings did not overlap in unsynchronized cells (Fig. 3E), indicating that CLUH is not a constitutive component of the centrosome. Taken together, our results show an increased formation of the CLUH-astrin-1 complex with the progression of the cell cycle, with highest levels in S and G2 phases, and a dissolution of the complex in mitosis.

Loss of astrin or kinastrin do not recapitulate CLUH-dependent mitochondrial phenotypes
What is the function of CLUH in complex with astrin-1 and kinastrin? We first asked whether loss of astrin or kinastrin phenocopies CLUH-dependent mitochondrial abnormalities. To avoid possible compensatory effects in stable KO clones, we employed inducible CRISPR-Cas9 astrin and kinastrin KO HeLa cells (hereafter referred to as SPAG5 and KNSTRN ind-KO) in which there is constant expression of sgRNAs complementary to the target genes and the expression of the Cas9 is doxycycline inducible (Kern et al, 2017;McKinley & Cheeseman, 2017). Using this system, we induced an acute deletion of astrin and kinastrin for four days, in many but not all cells (Figure 4-figure supplement 1A, B). We focused on previously reported phenotypes observed in CLUH KO cells or tissues, such as mitochondrial clustering, mitochondrial fragmentation, loss of mtDNA, and decreased assembled respiratory complexes (Gao et al, 2014;Schatton et al, 2017;Wakim et al, 2017). Mitochondrial distribution, morphology, ultrastructure and mtDNA levels were indistinguishable in WT, SPAG5 and These data indicate that CLUH controls mitochondrial function independent from the interaction with astrin-1 or kinastrin.

Loss of astrin rewires metabolic pathways to promote anabolism
To define how astrin and kinastrin affect cell function, we performed label free quantitative proteomics in SPAG5 and KNSTRN ind-KO cells. In line with the short time of induced KOs, only minor differences in the overall proteome in SPAG5 and KNSTRN ind-KO cells compared to ind-WT cells was observed ( Figure 5A and Supplementary File 3). Notably, CLUH levels were unaffected by the loss of its interactors ( Figure 5A and Supplementary File 3). Astrin depletion caused a slight decrease of proteins encoded by CLUH mRNA targets ( Figure 5A; marked in blue). Downregulated KEGG pathways, analyzed using the EnrichR webtool (using a generous cut-off of p≤0.05; q≤0.15) (Chen et al, 2013;Kuleshov et al, 2016;Xie et al, 2021), largely overlapped with metabolic pathways affected by CLUH deletion (TCA cycle, fatty acid and amino acid degradation; Figure 5B) (Schatton et al, 2017). Interestingly, metabolic pathways connected to nucleotide biosynthesis (purine and pyrimidine metabolism), the pentose phosphate pathway (PPP) and glycolysis were upregulated ( Figure 5C). Furthermore, mTORC1 appeared as one of the most enriched terms in pathway analysis of perturbed proteins The proteomic profile of SPAG5 ind-KO cells is consistent with a previous study that identified astrin as a negative regulator of mTORC1 (Thedieck et al, 2013). We confirmed that mTORC1 is hyperactive in SPAG5 ind-KOs upon starvation for 8 h in HBSS ( and sedoheptulose-7-phosphate, in agreement with high glycolytic rate and an increase in the PPP ( Figure 5D, E). In addition, the TCA cycle intermediates citrate, isocitrate and malate were increased ( Figure 5F). Both the TCA cycle and the PPP intermediates are crucial for synthesis of nucleotides, several of which accumulated in astrin-deficient cells ( Figure 5G). Amino acid levels were unaffected (Figure 5-figure supplement 2F), with the exception of an increase of ornithine, a precursor of polyamines, which were also more abundant ( Figure 5H). KNSTRN ind-KO cells did not show detectable metabolic differences to control cells ( Figure 5D-H and   Figure 6C). In addition, most amino acids that feed into the TCA cycle were decreased ( Figure   6D), including aspartate, which is mainly produced by oxaloacetate and is an important precursor of pyrimidines ( Figure 7D). Glutamine was instead elevated, as well as aketoglutarate ( Figure 7C, D). In addition, cells lacking CLUH showed a perturbation of polyamines metabolism, with a prominent reduction of spermidine, and decreased levels of some nucleotides, especially inosine monophosphate (IMP), the first nucleotide in the synthesis pathway of purines ( Figure 6E, F). Thus, upon starvation CLUH KO cells fail to maintain the TCA cycle intermediates (anaplerosis), and surprisingly show defects also in other anabolic pathways that are linked to cell growth and are regulated by mTORC1. This metabolic profile is strikingly opposite to that of SPAG5 ind-KO cells, and indicate that the mTORC1 hyperactivation in absence of CLUH is not reflected in the cellular metabolic rewiring. This data points to an important role of CLUH in the mTORC1-dependent rewiring of cellular metabolism, by supporting mitochondrial anaplerotic pathways.

CLUH controls cell cycle progression
In proliferating cells mitochondrial metabolism is constantly adapted to the specific metabolic needs of the different cell cycle phases. Metabolomic analyses during starvation (a treatment that arrests cell proliferation and blocks entry in G2/M) have revealed opposite signatures in cells lacking CLUH or astrin, indicating a perturbed metabolic rewiring. We therefore hypothesized that the functional significance of the interaction of CLUH with astrin is to couple mitochondrial metabolism to cell cycle progression. Assessment of unsynchronized cells using markers for different phases of the cell cycle showed a reduction of the levels of phosphorylated retinoblastoma (pRB1-Ser807/811) in absence of CLUH ( Figure 7A, B). pRB1 is hyperphosphorylated during the progression through the cell cycle by cyclin D/CDK4/6. This phosphorylation releases the RB inhibitory binding to E2F transcription factors, thus allowing gene transcription required for the progression through S phase.
To further investigate the cell cycle dysregulation in CLUH KO cells, we used a double thymidine block (DTB) to synchronize wildtype and CLUH KO HeLa cells at the beginning of Cyclin D3 also showed a flattener profile in the KO compared to wildtype cells during the first 8 h after release ( Figure 7H, I). In contrast, the inactivating phosphorylation of CDK1 on Tyr15, which occurs in G2 and allows progression into M, and the phosphorylation of serine 10 of histone 3, which marks cells in mitosis showed a similar temporal profile to wildtype cells ( Figure 7H, I). These results hint at the ability of CLUH KO cells to overcome cell cycle checkpoints at the G1/S transition. Remarkably, while CLUH levels did not change during the cell cycle, astrin-1 and to a lesser extent astrin-2 levels dropped prematurely in CLUH KO cells 6 h after synchronization, a time point where most cells are in G2 ( Figure 7H, I). This data is consistent with the increase of astrin-1 and its interaction with CLUH in G2 that we observed

DISCUSSION
We have uncovered a role of CLUH in integrating mitochondrial metabolism with cell cycle progression. Mechanistically, CLUH controls the synthesis and the stability of astrin-1, the full-length protein product of the SPAG5 gene. Astrin is a coiled-coil protein, with a wellestablished role during mitosis to stabilize kinetochore-microtubule interactions and allow correct orientation of chromosomes at the metaphase plate (Manning et al, 2010;Dunsch et al, 2011;Kern et al, 2017;Ying et al, 2020). Moreover, additional roles of astrin have been reported in interphase, including mTORC1 inhibition (Thedieck et al, 2013), centriole duplication (Kodani et al, 2015), and the recovery from DNA damage at the G2/M transition (Halim et al, 2013). We show here that the interaction of CLUH with astrin is regulated during the cell cycle, and ensures the matching of the mitochondrial metabolic output with the progression of the cell cycle.
SPAG5 upregulation in several cancers has been associated with poor prognosis (Yuan et al, 2014;Abdel-Fatah et al, 2016;Bertucci et al, 2016;Zhou et al, 2018;Li et al, 2019). Human cancer cell lines show the expression of two isoforms of astrin, however the origin of these isoforms has not been investigated up to now. We demonstrate that the two astrin isoforms arise by alternative translation initiation, and are regulated by the presence of an uORF and downstream AUGs with good Kozak sequences. Intriguingly, CLUH interacts only with the full-length astrin-1 protein. The N-terminal region of astrin is an unstructured domain that is dispensable for interaction with kinastrin, DYNLL1, MYCBP and the kinetochore (Kern et al, 2017). Recently, the PLK1 kinase has been shown to interact with and phosphorylate the astrin N-terminus, an important step to stabilize the kinetochore-microtubule attachment (Geraghty et al, 2021). Based on our findings, we speculate that N-terminal phosphorylations are responsible for the lack of interaction of CLUH and astrin-1 in prometaphase. Thus, astrin detaches from CLUH during mitosis when it is required for metaphase to anaphase progression.
Interestingly, CLUH appears to regulate astrin-1 expression at multiple levels. The SPAG5 transcript represents an exception to the finding that mRNAs bound by CLUH mainly encode mitochondrial proteins (Gao et al, 2014). We show here that the SPAG5 mRNA decays faster and that the synthesis of astrin-1 is impaired in absence of CLUH. It is known that during cytokinesis the E3-ubiquitin ligase MID2 ubiquitinates astrin, targeting it for proteosomal degradation (Gholkar et al, 2016). CLUH may be implicated in supporting synthesis of astrin-1 to replenish its levels in G1. Astrin-1 then accumulates during the S and G2 phases of the cell cycle and is stabilized, together with kinastrin, by binding CLUH. While we established the role of CLUH binding to prevent astrin-1 degradation, more experiments are required to determine to what extent the synthesis of astrin-2 is also modulated by CLUH.
The interaction of CLUH with astrin-1 led us first to investigate the possibility of a role of astrin-1 in concert with CLUH in regulating mitochondrial metabolism. Our data do not support this scenario, since depletion of astrin or kinastrin do not recapitulate the prominent mitochondrial defects observed upon CLUH downregulation or knock-out. Upon loss of astrin or kinastrin, mitochondria are normally dispersed in the cytoplasm, show normal morphology and ultrastructure, do not lose mtDNA and efficiently assemble respiratory supercomplexes.
SPAG5 ind-KO cells showed a mild decrease of the levels of mitochondrial proteins encoded by CLUH target mRNAs. Furthermore, starved cells lacking astrin or CLUH display strikingly opposite metabolic profiles. Finally, astrin and kinastrin are not detected in polysomal fractions, in contrast to CLUH. In agreement with our data, a recent study has also identified astrin as a CLUH partner and showed that SPAG5 knock-down does not affect the interaction of CLUH with nuclear-encoded mitochondrial proteins occurring before their import in the organelle and revealed by Bio-ID proximity labelling (Hemono et al, 2022). It is possible that the formation of the astrin-1-CLUH complex competes with the function of CLUH at the polysomes. This remains a hypothesis at present and more work will be required to assess this possibility.
Our data establish an essential role of CLUH in the metabolic rewiring directed by mTORC1 signaling. Upon starvation, both CLUH KO and SPAG5 ind-KO cells showed inappropriate mTORC1 activation, in agreement with astrin being a negative regulator of mTORC1 (Thedieck et al, 2013). It remains to be established if mTORC1 hyperactivation observed upon loss of CLUH is entirely dependent on astrin destabilization, or whether different pathways are at play. However, the metabolic defects caused by loss of CLUH were discordant with this signaling, with a decrease in the TCA cycle intermediates and in the pool of glycolytic intermediates, some nucleotides, and polyamines. Enzymes of the TCA cycle as well as characteristic anaplerotic enzymes, such as pyruvate carboxylase, proprionyl-CoA carboxylase and other branched amino-acid catabolic enzymes are encoded by bona fide CLUH target mRNAs (Gao et al, 2014). These pathways are essential to replenish the TCA cycle intermediates to compensate for their loss upon activation of biosynthetic pathways following mTORC1 signaling. Indeed, a compensatory increase in anaplerotic pathways occurs under conditions of chronic mTORC1 activation (Dutchak et al, 2018). Therefore, CLUH tunes mTORC1 signaling in two independent ways, by regulating the expression of astrin-1 and by ensuring the maintenance of mitochondrial anaplerotic pathways via its RNA-binding function Proliferating cells must shunt metabolites into mitochondrial pathways that promote cell growth or oxidative respiration and ATP production, depending on the cell cycle phase (DeBerardinis et al, 2008;Salazar-Roa & Malumbres, 2017). In G1 cells double their mass and depend on glycolysis to fuel the TCA cycle and sustain biosynthetic pathways, while in S synthesis of nucleotides is required and glutamine oxidation is prevalent. Interestingly, the TCA cycle intermediates fluctuate during the cell cycle, and it is known that these variations are not transcriptionally regulated (Olsen et al, 2010;Ahn et al, 2017). In mammalian cells, mTORC1 activation is important for the transition from G1 to S, an important check point when the energy levels and the growth status of the cell are sensed, and from G2 to M (Cuyàs et al, 2014). We hypothesize that by regulating astrin-1 and stabilizing an RNA regulon involved in the TCA cycle and anaplerotic pathways (Gao et al, 2014) Astrin is a component of centriolar satellites in interphase, and it is required for the centrosomal localization of CDK5RAP2, a protein involved in microcephaly, and for proper centriole duplication in S phase (Thein et al, 2007;Kodani et al, 2015), a step which is crucial for cell cycle progression. Astrin-1 overexpression depletes endogenous CLUH from the cytoplasm and recruits it to structures that are close to centrosomes or contain centrosomal markers.
However, we could not demonstrate the recruitment of endogenous CLUH at the centrosome at least in unsynchronized cells. Whether and how astrin-1 is involved in regulating CLUH subcellular localization remains to be determined.
Finally, our data add to the evidence that astrin plays additional roles outside mitosis. Astrin was also identified in an unbiased screen to be essential for the reentry in mitosis after DNA damage (Halim et al, 2018). This interphase role of astrin should be further explored in view of our findings, as possibly involving dysregulated or enhanced CLUH activity. HeLa cells, as many human cancer cells, overexpress not only astrin-1 but also astrin-2, an isoform of astrin competent to bind all other interactors (kinastrin, DYLNN1 and MYCPB) and the kinetochore (Kern et al, 2017), but not CLUH. We propose that expression of astrin-2 endows cancer cells with the ability to proliferate, despite a dysregulated coupling between mitochondrial metabolism and the cell cycle, making CLUH a possible novel therapeutic target in cancers overexpressing SPAG5. In conclusions, our findings reveal a novel post-transcriptional mechanism coordinating mitochondrial metabolism and the cell cycle.
HeLa cells have not been further validated, but in all cases parental HeLa cells have been used as controls. Immortalized WT and Cluh KO MEFs were previously described (Gao et al, 2014).
HEK293T cells were purchased from a vendor several years ago and not further validated.
These cells were used to stably express inducible 3xFLAG-kinastrin and employed only for a biochemical experiment which was confirmed with independent cell lines. To this end, full length N-terminally 3xFLAG tagged human KNSTRN ORF was cloned into pTREx-DEST30 vector by serial Gateway recombinations (Invitrogen) according to the manual. Stable transfected HEK293T cells were generated with the Flp-In T-REx 293 cell system (Invitrogen) following the manual. Positive cells were selected with media containing 1.5 mg/mL hygromycin (InvivoGen) and 150 µg/mL blasticidin (InvivoGen) starting 24 h after transfection and single colonies were picked to generate a monoclonal cell line. All cell lines were regularly checked for Mycoplasma contamination and found negative.

Cloning and mutagenesis of SPAG5 constructs
The underlying sequence of human SPAG5 used in this study can be found under the accession number NM_006461.3 on the NCBI database. Human full length SPAG5 ORF (ATG1-SPAG5) or excluding the first 453 nucleotides (Δ151-SPAG5) were cloned into p3xFLAG-CMV-14 (Sigma-Aldrich) using NotI or NotI/ClaI restriction sites, respectively. To obtain a construct including the 5´ UTR (5UTR-SPAG5), the 5´ UTR of SPAG5 was cloned into Δ151-SPAG5-FLAG using HindIII restriction site. To mutagenize ATGs to GGG, DpnI site directed mutagenesis was employed. Mutagenized base pair positions and construct names are indicated in Figure 1C and S1A.

Cell transfection
Cells were transfected with Lipofectamine 2000 (Invitrogen) in a plasmid to transfection reagent ratio of 1:5 according to the manual. After 24 h of overexpression, cells were harvested by scraping and pelleting for lysis or fixed with respective reagent for immunofluorescence.
RNA interference of CLUH was done as described before (Gao et al, 2014). Briefly, HeLa cells were transfected with 100 nM of siRNA against human CLUH or control siRNA using

Synchronization of cells
Cells were synchronized by double thymidine block (DTB) using an adapted protocol (Dai et al, 2018). Briefly, cells were plated in desired amount and treated the next day with 2 mM thymidine (Sigma-Aldrich) for 23 h, afterwards cells were released in standard media without thymidine for 10 h and second thymidine block (2 mM thymidine) was performed for 13.5 h.
Next day, cells were released and collected at indicated time points depending on the experiment (Figure 3-figure supplement 1A, Figure 7C). To enrich cells in PM phase, cells were treated with 100 ng/µl nocodazole (Sigma-Aldrich) after 3 h of release and collected after 12 additional hours. To block cells in G0/G1, cells were starved in media without serum for 16 h.
Samples were stored at -20°C or run immediately on SDS-PAGE.

Sample preparation for mass spectrometry
For IP of CLUH followed by mass spectrometry of WT cells enriched in G2, experiments were carried out as described before using 400 µg lysate as input and elution was done with 30 µl SP3 lysis buffer (5% SDS in 1x PBS) by vortexing for 1 min and boiling at 95°C for 5 min.
Afterwards, proteins were reduced with 5 mM dithiothreitol for 30 min at 55°C and alkylated with 40 mM chloroacetamide at RT for 30 min in the dark. Next, samples were centrifuged at 20,000 x g for 10 min and supernatant was transferred to new tube and stored at -20°C before mass spectrometry was performed. For SILAC labelled samples, beads were resuspended in 50 µl of elution buffer 1 (2M urea, 50 mM triethylammoniumbicarbonate, 1 mM DTT, 5 ng/µl trypsin) and incubated at RT for 30 min while shaking. Beads were centrifuged and supernatant transferred to a new tube. Beads were washed twice with elution buffer 2 (2M urea, 50 mM triethylammoniumbicarbonate, 5 mM chloroacetamide) and centrifuged. The eluates were combined. Next, proteins were digested with lysyl endopeptidase (Wako Pure Chemical Industries) and trypsin (Sigma-Aldrich) with an enzyme:substrate ratio of 1:75 at 37 °C for 16 h. Next day, samples were acidified with formic acid to stop enzymatic digestion, purified with and loaded on StageTips as described before (Rappsilber et al, 2007).
HeLa WT, SPAG5 and KNSTRN ind-KO cells were collected by scraping, pelleted and resuspended in appropriate amount of lysis buffer [50 mM Tris-HCl, pH 7.4; 150 mM NaCl; 1 mM EDTA, pH8; 1% IGEPAL CA-630; 0.25% sodium deoxycholate freshly supplemented with protease inhibitor cocktail (Sigma-Aldrich)]. Lysates were passed 3x through syringe (30G x 1/2", B. Braun Sterican) and incubated on ice for 30 min. After centrifugation for 30 min at 20,000 x g at 4°C, protein amounts were determined using standard Bradford assay (Biorad) and 30 µg of protein lysate were precipitated with acetone. Briefly, 4x volume of ice-cold acetone were added to lysates, incubated for 15 min at -80°C followed by 90 min incubation at -20°C and centrifugation for 15 min at 16,000 x g. Pellets were washed in ice-cold acetone, airdried and resuspended in 50 µl of 8 M urea in 50 mM triethylammoniumbicarbonate including protease inhibitor cocktail (Roche). Afterwards proteins were reduced with 5 mM dithiothreitol for 1 h at 25°C and alkylated with 40 mM chloroacetamide for 30 min in the dark. Next, proteins were digested with lysyl endopeptidase (Wako Pure Chemical Industries) with an enzyme:substrate ratio of 1:75 at 25°C for 4 h. Samples were diluted with 50 mM triethylammoniumbicarbonate to reach a final urea concentration of 2 M. Then proteins were digested with trypsin (Sigma-Aldrich) with an enzyme:substrate ratio of 1:75 and incubation at 25°C for 16 h. Next day, samples were acidified with formic acid to stop enzymatic digestion, purified with and loaded on StageTips as described before (Rappsilber et al, 2007). Samples were stored in dried StageTips at 4°C until mass spectrometry was performed.

Mass spectrometry of CLUH IP after SILAC
SILAC labelled samples were analyzed on a Q Exactive Plus Orbitrap (Thermo Scientific) mass spectrometer that was coupled to an EASY nLC (Thermo Scientific). Peptides were loaded with solvent A (0.1% formic acid in water) onto an in-house packed analytical column (50 cm x 75 µm I.D., filled with 2.7 µm Poroshell EC120 C18, Agilent). Peptides were chromatographically separated at a constant flow rate of 250 nL/min using the following gradient: 7-23% solvent B (0.1% formic acid in 80 % acetonitrile) within 35.0 min, 23-32% solvent B within 5.0 min, 32-85% solvent B within 5.0 min, followed by washing and column equilibration. The mass spectrometer was operated in data-dependent acquisition mode. The MS1 survey scan was acquired from 300-1750 m/z at a resolution of 70,000. The top 10 most abundant peptides were isolated within a 1.8 Th window and subjected to HCD fragmentation at a normalized collision energy of 27%. The AGC target was set to 5e5 charges, allowing a maximum injection time of 108 ms. Product ions were detected in the Orbitrap at a resolution of 35,000. Precursors were dynamically excluded for 20.0 s. All mass spectrometric raw data were processed with Maxquant (version 1.5.3.8) using default parameters. Briefly, MS2 spectra were searched against the Uniprot HUMANc_UP000005640.fasta (downloaded at: 26.08.2020) database, including a list of common contaminants. False discovery rates on protein and PSM level were estimated by the target-decoy approach to 1% (Protein FDR) and 1% (PSM FDR) respectively.
The minimal peptide length was set to 7 amino acids and carbamidomethylation at cysteine residues was considered as a fixed modification. Oxidation (M) and Acetyl (Protein N-term) were included as variable modifications. SILAC/dimethyl labeling quantification was used, and the re-quantify option was enabled.
The MS1 survey scan was acquired from 300 to 1750 m/z at a resolution of 70,000. The top 10 most abundant peptides were isolated within a 1.8 Th window and subjected to HCD fragmentation at a normalized collision energy of 27%. The AGC target was set to 5e5 charges, allowing a maximum injection time of 110 ms. Product ions were detected in the Orbitrap at a resolution of 35,000. Precursors were dynamically excluded for 10 s. All mass spectrometric raw data were processed with MaxQuant version 1.5.3.8 (Tyanova et al, 2016b) using default parameters. Briefly, MS2 spectra were searched against a canonical Uniprot human fasta database, which was modified by replacing the default entry for SPAG5 (Q96R06) by two separate entries representing (i) the N-terminal 125 amino acids and (ii) the C-terminal sequence from position 126 on. The MaxQuant default list was used to filter for common contaminants. False discovery rates on protein and PSM level were estimated by the targetdecoy approach to 1% (Protein FDR) and 1% (PSM FDR) respectively. The minimal peptide length was set to 7 amino acids and carbamidomethylation at cysteine residues was considered as a fixed modification. Oxidation (M) and Acetyl (Protein N-term) were included as variable modifications. The match-between runs option was restricted to replicates of the same condition. LFQ quantification was used with default settings. LFQ intensities were loaded into in Perseus version 1.6.1.1 (Tyanova et al, 2016a). Decoys and potential contaminants were removed and the dataset was filtered for at least 4 out of 4 values in at least one condition.
Remaining missing values were imputed with random values from the left end of the intensity distribution using Perseus defaults. Two sample Student's T-test were calculated using permutation-based FDR estimation.

Mass spectrometry of SPAG5 and KNSTRN ind-KO cells
For proteomics of HeLa WT ind, SPAG5 and KNSTRN ind-KO cells, peptide digests were analyzed on a Q Exactive plus Orbitrap (Thermo Scientific) mass spectrometer that was coupled to an EASY nLC (Thermo Scientific). Samples were loaded onto an in-house packed analytical column (50 cm x 75 µm I.D., filled with 2.7 µm Poroshell EC120 C18, Agilent). Peptides were separated at a flow rate of 250 nL/min and the following gradient: 3-5% solvent B (0.1% formic acid in 80 % acetonitrile) within 1.0 min, 5-30% solvent B within 91.0 min, 30-50% solvent B within 17.0 min, 50-95% solvent B within 1.0 min, followed by washing with 95 % solvent B for 10 min. DDA runs for spectrum library generation were acquired from distinct pools of the sample groups and Hek293 cell digests fractionated high pH HPLC. MS1 survey scan were acquired at a resolution of 70,000. The top 10 most abundant peptides were isolated within a 2.0 Th window and subjected to HCD fragmentation with normalized collision energy of 27%.
The AGC target was set to 5e5 charges, allowing a maximum injection time of 105 ms. Product ions were detected in the orbitrap at a resolution of 35,000. Precursors were dynamically excluded for 20.0 s. Sample runs were acquired in data-independent mode using 10 variable windows covering the mass range from m/z 450 to m/z 1200. MS1 scans were acquired at 140,000 resolution, maximum IT restricted to 120 ms and an AGC target set to 5e6 charges.
The settings for MS2 scans were 17,500 resolution, maximum IT restricted to 60 ms and AGC target set to 5e6 charges. The default charge state for the MS2 was set to 4. Stepped normalized collision energy was set to 23.5, 26 and 28.5. All spectra were acquired in profile mode. A hybrid spectrum library was generated in Spectronaut 13 (Bruderer et al, 2015) using DDA library runs, DIA sample runs and a canonical human sequence file (SwissProt, 20416 entries) downloaded from Uniprot. Spectronaut default settings were used for the analysis of the DIA runs. Protein identifications were filtered for q-values below 0.01 and normalized intensities were exported for subsequent statistical analysis in Perseus 1.6.1.1 (Tyanova et al, 2016b).
Intensities were transformed to log2 values and the dataset was filtered for at least 4 out of 4 values in at least one condition. Remaining missing values were imputed with random values from the left end of the intensity distribution (with 0.3 sd, downshift 2 sd). Two sample Student´s T-tests were calculated using permutation based FDR estimation.

Proteomics visualization and pathway analysis
Enriched proteins of IP experiment or proteomics results were visualized as volcano plots using Instant Clue software (Nolte et al, 2018) and pathway analysis was carried out using the EnrichR webtool with a cut off of q≤0.05 and log2 fold change ≥ 3 for the IP and a cut off of q≤0.15 and p≤0.05 for the proteomics analysis (Chen et al, 2013;Kuleshov et al, 2016;Xie et al, 2021).

Isolation of mitochondria
Cells were collected from confluent 15 cm plates with trypsinization, washed twice with PBS, and resuspended in an ice-cold mitochondria isolation buffer containing 20 mM HEPES, pH 7.6; 220 mM mannitol; 70 mM sucrose; 1 mM EDTA; 0.2% fatty acid-free bovine serum albumin (BSA). After 20 min of incubation on ice, cells were homogenized using the rotational engine homogenizer (Potter S, Sartorius; 30 strokes, 1200 rpm) followed by centrifugation at 850 x g. Next, mitochondria were pelleted at 8,500 x g for 10 min at 4 °C, washed with BSAfree buffer, and protein concentration was determined with Bradford reagent (Sigma-Aldrich).
For further analysis, mitochondria were subjected to blue native polyacrylamide gel electrophoresis (BN-PAGE) followed by western blotting or determination of the ingel activity of respiratory complexes

Analysis of mitochondrial respiratory complexes with BN-PAGE
20 mg of mitochondria were lysed with digitonin (Calbiochem; 6.6 g/g protein) for 15 min on ice with occasional vortexing and cleared from insoluble material for 20 min at 20,000 x g at 4 °C. Lysates were combined with Coomassie G-250 (0.25% final concentration). Mitochondrial respiratory supercomplexes were resolved with BN-PAGE using the 4-16% NativePAGE Novex Bis-Tris Mini Gels (Invitrogen) in a Bis-Tris/Tricine buffering system with cathode buffer initially supplemented with 0.02% G-250 followed by the 0.002% G-250. For complex I in-gel activity, gels were incubated at RT in a buffer containing 0.01 mg/mL NADH and 2.5 mg/mL nitrotetrazolium blue in 5 mM Tris-HCl, pH 7.4.

Sample collection to measure anionic metabolites, amino acids and polyamines
WT, SPAG5 and KNSTRN ind-KO cells were induced as described before and 1,000,000 cells were plated in 6 well plates the day before the extraction. Next day, the cells were either collected immediately (basal condition) or starved for 8 h in HBSS media (#14025092, Gibco) in absence of doxycycline. Cells were washed twice with buffer containing 75 mM ammonium carbonate (pH7.4) and metabolites were extracted with cold (-20°C) extraction solvent (40:40:20 acetonitrile:methanol:water) and incubation for 10 min at -20°C. Supernatant was collected and extraction was repeated. Afterwards, cells were scraped on ice and combined with the supernatant of the previous step. Samples were immediately dried in a speed vac concentrator and dried pellets were kept at -80°C until mass spectrometry was performed.

Anion-Exchange Chromatography Mass Spectrometry (AEX-MS) for the analysis of anionic metabolites
Extracted metabolites were re-suspended in 150 µl of Optima LC/MS grade water (Thermo Fisher Scientific), of which 100 µl were transferred to polypropylene autosampler vials (Chromatography Accessories Trott, Germany) before AEX MS analysis. The samples were analysed using a Dionex ion chromatography system (Integrion, Thermo Fisher Scientific) as described previously (Schwaiger et al, 2017). In brief, 5 µL of polar metabolite extract were injected in push partial mode using an overfill factor of 3, onto a Dionex IonPac AS11-HC column (2 mm × 250 mm, 4 µm particle size, Thermo Fisher Scientific) equipped with a Dionex IonPac AG11-HC guard column (2 mm × 50 mm, 4 µm, Thermo Fisher Scientific). The column temperature was held at 30°C, while the auto sampler was set to 6°C. A potassium hydroxide gradient was generated using a potassium hydroxide cartridge (Eluent Generator, Thermo Scientific), which was supplied with deionized water. The metabolite separation was carried at a flow rate of 380 µL/min, applying the following gradient conditions: 0-3 min, 10 mM KOH; Areas of the cellular pool sizes were normalized to the internal standards (citric acid D4), which were added to the extraction buffer, followed by a normalization to the protein content of the analyzed sample. One sample of CLUH KO and one of SPAG5 ind-KO cells upon HBSS starvation has been classified as outlier and removed from analysis. Samples were classified as outliers due to PCA plot and tremendous drift of measured values from other samples.

LC-MS analysis of cellular pool sizes of amino acids and polyamines
For amino acid analysis, the benzoylchlorid derivatization method (Wong et al, 2016) was used.
In brief, 20 µl of the polar phase of each sample, were mixed with 10 µl of 100 mM sodium carbonate (Sigma-Aldrich) followed by the addition of 10 µl 2% benzoylchloride (Sigma-Aldrich) in acetonitrile (VWR). Samples were analyzed using an Acquity iClass UPLC (Waters) connected to a Q-Exactive HF (Thermo Fisher Scientific). For analysis, 1 µl of the derivatized sample was injected onto a 100 x 1.0 mm HSS T3 column, packed with 1.8 µm particles (Waters). The flow rate was 100 µL/min and the buffer system consisted of buffer A (10 mM ammonium formate, 0.15% formic acid in water) and buffer B (acetonitrile). The gradient was: 0% B at 0 min; 0-15% B 0-0.1 min; 15-17% B 0.1-0.5 min; 17-55% B 0.5-14 min, 55-70% B 14-14.5 min; 70-100% B 14.5-18 min; 100% B 18-19 min; 100-0% B 19-19.1 min, 19.1-28 min 0% B. The mass spectrometer was operating in positive ionization mode monitoring and the mass range was set to m/z 50-750. The heated ESI source settings of the mass spectrometer were: Spray voltage 3.5kV, capillary temperature 275°C, sheath gas flow 40 AU and aux gas flow 20 AU at a temperature of 300°C. The S-lens was set to 60 AU. Data analysis was performed using the TraceFinder software (Version 4.1, Thermo Fisher Scientific). Identity of each compound was validated by authentic reference compounds, which were injected and analyzed independently. Extracted ion chromatograms (XIC) were extracted as [M + H] + ions with a mass accuracy (<5 ppm). Areas of the cellular pool sizes of the analyzed amines were normalized to their corresponding 13 C 15 N internal standard or, if no corresponding 13 C 15 N compound was present, they were normalized to the 13 C 15 N leucine. Following the normalization to the internal standard the values were normalized to the protein content of the analyzed sample.

Measurement of RNA stability
To measure mRNA stability, Click-iT Nascent RNA Capture Kit (Invitrogen) was used as described before (Schatton et al., 2017). Briefly, 500,000 cells were seeded on 3.5 cm dishes.
The following day, endogenous RNA was labeled with 0.2 mM 5-ethynyl uridine (EU) for 24 h and collected either immediately (0 h time point) or after 8 h incubation with media w/o EU with Trizol reagent (Invitrogen). Total RNA was isolated according to the instructions of the Trizol reagent manual and 3 µg RNA were biotinylated with biotin-azide using Click-iT chemistry reaction. Afterwards, RNA was precipitated with glycogen, 7.5 M ammonium acetate and 100% ice cold ethanol for 16 h and centrifugation at 13,000 x g for 20 min at 4°C.
RNA was washed twice with 70% ethanol, air-dried and resuspended in distilled H2O. 1.5 µg RNA were incubated with prewashed Dynabeads MyOne Streptavidin T1 beads (Invitrogen) for 30 min in presence of RNaseOUT (Invitrogen). Afterwards beads were washed several times and RNA bound on beads was retrotranscribed with SuperScript VILO cDNA synthesis kit (Invitrogen) according to the manual. cDNA was stored at -20°C until qRT-PCR was performed. mRNA half-lives were calculated as described before (Schatton et al, 2017).

RNA isolation, cDNA synthesis, DNA isolation and quantitative real-time PCR
RNA was isolated with Trizol reagent (Invitrogen) according to the manual. 2 µg of total RNA were retro-transcribed using the SuperScript First-Strand Synthesis System (Invitrogen) with random hexamer primers according to the instructions of the manual. To isolate genomic DNA, cells were lysed with digestion buffer [100 mM NaCl; 10 mM Tris-HCl, pH8; 25 mM EDTA, pH8; 0.5% SDS supplemented freshly with 0.1 mg/mL proteinase K (Roche)] at 55°C for 16 h.

Immunofluorescence and transmission electron microscopy
Cells were seeded in an appropriate amount on coverslips the day before. Next day, cells were washed twice with 1x PBS and fixed for 15 min with 4% PFA/PBS (pH7.4) at RT. Afterwards, cells were permeabilized with 0.2% Triton X-100 in 1x PBS for 10 min and blocked in 10% pig or goat serum for 10 min. Then coverslips were stained with primary antibodies diluted in 1% pig or goat serum in 1x PBS for 2h at RT or for 16h at 4°C. The following primary antibodies were used: rabbit polyclonal anti-CLUH antibody ( DNA dye) and mounted with Fluorsave reagent (Calbiochem). Transmission electron microscopy on cells was performed as previously described (Gao et al, 2014).

Microscopy and image analysis
Immunofluorescence images were acquired with a spinning-disk confocal microscope (UltraVIEW VoX, PerkinElmer) using a 60x objective. Images represent a single plane and were deconvoluted using ImageJ (NIH) and brightness was adjusted equally in the entire images. Specificity of the anti-CLUH antibody has been proven before (Pla-Martin et al, 2020).
For mitochondrial morphology assessment, a half-automated macro for ImageJ (Mitomorph) has been employed (Yim et al, 2020). At least 66 (basal) or 44 (galactose) cells per genotype per experiment were analyzed. Micrographs were acquired on a Jeol Jem2100Plus electron microscope operating at a voltage of 120 V using a GATAN OneView camera.

Propidium iodide staining and flow cytometry
Cells were synchronized as described before ( Figure 8A)